You are on page 1of 78

Linkping University Medical Dissertations

No 706

Radiation Dose
and Image Quality
in Diagnostic
Radiology

Optimization of the dose


image quality relationship
with clinical experience
from scoliosis radiography,
coronary intervention and a
flat-panel digital detector

Hkan Geijer

Department of Radiology, rebro University Hospital, SE-701 85 rebro


and
Department of Medicine and Care
Faculty of Health Sciences, Linkping University, SE-581 85 Linkping

Linkping and rebro 2001


Linkping University Medical Dissertations
No 706

Radiation Dose andImage Quality in


Diagnostic Radiology

Optimization of the dose image quality relationship with clinical


experience from scoliosis radiography, coronary intervention and a
flat-panel digital detector

Hkan Geijer

Department of Radiology, rebro University Hospital, SE-701 85 rebro


and
Department of Medicine and Care
Faculty of Health Sciences, Linkping University, SE-581 85 Linkping

Linkping and rebro 2001


Cover art from Humour noir et hommes en blanc by Claude Serre 1975
by permission from Editions Glnat, France.

ISBN 91-7373-143-9
ISSN 0345-0082
Copyright Hkan Geijer (pp 1-76)
Printed in Sweden
Linkpings Tryckeri AB

Linkping November 2001


To Mia

denna skrift handlar om genomforskning


Contents

Contents

1 ABSTRACT................................
................................
.........................7

2 SAMMANFATTNING................................
................................
...........8

3 PRELIMINARY REPORTS................................
................................
... 9

4 ORIGINAL PAPERS................................
................................
...........10

5 ABBREVIATIONS................................
................................
..............11

6 INTRODUCTION................................
................................
...............13
6.1 GENERAL INTRODUCTION....................................................................... 13
6.2 RADIATION DOSE ................................................................................. 14
6.3 IMAGE QUALITY ................................................................................... 14
6.4 THE CONCEPT OF OPTIMIZATION .............................................................. 16
6.5 SCOLIOSIS RADIOGRAPHY ...................................................................... 17
6.6 FLAT-PANEL DETECTOR ......................................................................... 18
6.7 CORONARY INTERVENTION ..................................................................... 18
7 AIMS ................................
................................
...............................
19

8 MATERIALS AND METHOD


S ................................
............................
20
8.1 RADIATION DOSE ................................................................................. 20
8.1.1 Entrance dose and ESD................................................................... 20
8.1.2 Absorbed dose from scattered radiation.......................................... 21
8.1.3 Kerma area-product......................................................................... 21
8.1.4 Effective dose................................................................................... 22
8.1.5 Alderson phantom............................................................................ 23
8.1.6 Monte Carlo simulation .................................................................... 24
8.1.7 Equivalent diameter and Reference Man......................................... 24
8.2 IMAGE QUALITY ................................................................................... 25
8.2.1 Image viewing conditions ................................................................ 25
8.2.2 Contrast-detail phantoms ................................................................ 25
8.2.3 Anthropomorphic phantoms............................................................. 27
8.2.4 Visual Grading Analysis (VGA) ........................................................ 28
8.3 SCOLIOSIS RADIOGRAPHY ...................................................................... 28
8.3.1 Radiographic techniques.................................................................. 28
8.3.2 Dose and image quality assessment ............................................... 29
8.3.3 Angle measurements ....................................................................... 31
8.3.4 Optimization of the digital exposure protocol ................................... 32
8.4 FLAT-PANEL DETECTOR ......................................................................... 32
8.4.1 Radiographic techniques.................................................................. 32
8.4.2 Dose and image quality assessment ............................................... 33
8.5 CORONARY INTERVENTION ..................................................................... 33
8.5.1 Radiographic techniques.................................................................. 33
8.5.2 Dose measurements ........................................................................ 34

-5-
Hkan Geijer

8.5.3 Image quality assessment ............................................................... 36


8.5.4 Optimization of settings ................................................................... 36
8.5.5 Clinical study................................................................................... 36
8.6 STATISTICAL METHODS ......................................................................... 37
8.7 ETHICAL CONSIDERATIONS..................................................................... 38
9 RESULTS................................
................................
.........................3 9
9.1 SCOLIOSIS RADIOGRAPHY ...................................................................... 39
9.1.1 Radiation dose................................................................................. 39
9.1.2 Image quality ................................................................................... 42
9.1.3 Angle measurements ....................................................................... 43
9.2 FLAT-PANEL DETECTOR ......................................................................... 44
9.2.1 Radiation dose and image quality ................................................... 44
9.3 CORONARY INTERVENTION ..................................................................... 46
9.3.1 Radiation dose................................................................................. 46
9.3.2 Image quality ................................................................................... 48
9.3.3 Patient dose data............................................................................. 49
10 DISCUSSION................................
................................
.................53
10.1 SCOLIOSIS RADIOGRAPHY ...................................................................... 53
10.2 FLAT-PANEL DETECTOR ........................................................................ 54
10.3 CORONARY INTERVENTION ..................................................................... 55
10.4 STATISTICAL METHODS ......................................................................... 58
10.5 RADIATION DOSE ................................................................................. 59
10.6 IMAGE QUALITY ................................................................................... 60
10.7 OPTIMIZATION METHODS ....................................................................... 63
10.7.1 General considerations .................................................................... 63
10.7.2 A practical approach to optimization ............................................... 65
11 CONCLUSIONS................................
................................
..............66

12 ACKNOWLEDGEMENTS................................
................................
. 67

13 REFERENCES................................
................................
...............68

-6-
Abstract

1 Abstract
X-rays are known to cause malignancies, skin damage and other side effects
and they are thus potentially dangerous. Therefore, it is essential and in fact
mandatory to reduce the radiation dose in diagnostic radiology as far as possi-
ble. This is also known as the ALARA (As Low As Reasonably Achievable) prin-
ciple. However, the dose is linked to image quality and the image quality may
not be lowered so far that it jeopardizes the diagnostic outcome of a radio-
graphic procedure. The process of reaching this balance between dose and im-
age quality is called optimization. The aim of this thesis is to propose and
evaluate methods for optimizing the radiation dose image quality relationship
in diagnostic radiography with a focus on clinical usefulness. The work was
performed in three main parts.
Optimization of scoliosis radiography: In the first part, two recently devel-
oped methods for digital scoliosis radiography (digital exposure and pulsed
fluoroscopy) were evaluated and compared to the standard screen-film method.
Radiation dose was measured as Kerma area-product (KAP), Entrance surface
dose (ESD) and effective dose; image quality was assessed with a contrast-
detail phantom and through Visual grading analysis. Accuracy in angle meas-
urements was also evaluated. The radiation dose for digital exposure was
nearly twice as high as the screen-film method at a comparable image quality
while the dose for pulsed fluoroscopy was very low but with a considerably
lower image quality. The variability in angle measurements was sufficiently low
for all methods. Then, the digital exposure protocol was optimized to a consid-
erably lower dose with a slightly lower image quality compared to the baseline.
Flat-panel detector: In the second part, an amorphous-silicon direct digital
flat-panel detector was evaluated using a contrast-detail phantom, measuring
dose as entrance dose. The flat-panel detector yielded a superior image quality
at a lower dose than both storage phosphor plates and screen-film. Equivalent
image quality compared to storage phosphor plates was reached at about one
third of the dose.
Optimization of percutaneous coronary intervention (PCI): In the third part,
influence of various settings on radiation dose and image quality in coronary
catheterisation and PCI was investigated. Based on these findings, the dose
rate for fluoroscopy was reduced to one-third. The dose reduction was evalu-
ated in a clinical series of 154 PCI procedures before and 138 after the optimi-
zation. Through this optimization, the total KAP was significantly reduced to
two-thirds of the original value.
In summary, this thesis indicates the possibility of dose reduction in diag-
nostic radiology through optimization of the radiographic process.

-7-
Hkan Geijer

2 Sammanfattning
Det r knt att rntgenstrlning kan orsaka cancersjukdomar, hudskador och
andra sidoeffekter. Drfr r det viktigt och ven freskrivet i lag att strlexpo-
neringen inom diagnostisk radiologi skall snkas s lngt som mjligt. Detta
kallas p engelska ALARA-principen (As Low As Reasonably Achievable). Strl-
dosen r kopplad till bildkvalitet och denna fr inte snkas s lngt att det
diagnostiska vrdet av en underskning ventyras. Processen att n en sdan
balans mellan dos och bildkvalitet kallas optimering. Syftet med denna av-
handling r att finna och utvrdera metoder fr att optimera frhllandet mel-
lan strldos och bildkvalitet inom diagnostisk radiologi med fokus p klinisk
anvndbarhet. Arbetet utfrdes i tre huvuddelar.
Optimering av skoliosrntgen: I frsta delen utvrderades tv nyligen ut-
vecklade metoder fr digital skoliosrntgen (digital exponering och pulsad
genomlysning). De jmfrdes ven med film-skrmsystem som var den tidigare
standardmetoden. Strldosen mttes som Kerma area-produkt (KAP), ytdos
(Entrance surface dose, ESD) och effektiv dos; bildkvaliteten vrderades med
ett kontrast-detaljfantom och genom visuell analys p kliniska bilder. Nog-
grannheten i vinkelmtningar vrderades ocks. Strldosen fr digital expone-
ring var nstan dubbelt s hg som fr film med jmfrbar bildkvalitet medan
pulsad genomlysning hade en mycket lg dos men betydligt smre bildkvalitet.
Variabiliteten i vinkelmtningar var tillrckligt lg i alla metoder. Drefter op-
timerades instllningarna fr digital exponering till en betydligt lgre strldos
med viss snkning av bildkvaliteten jmfrt med utgngslget.
Direktdigital detektor: I den andra delstudien utvrderades en direktdigital
detektor med ett kontrast-detaljfantom dr strldosen mttes som ingngsdos i
fantomet. Den direktdigitala detektorn gav bttre bildkvalitet vid lgre dos
jmfrt med bde bildplattor och film. Jmfrbar bildkvalitet med bildplattor
nddes vid ungefr en tredjedel av dosen.
Optimering av perkutan koronar intervention (PCI): I tredje delstudien un-
dersktes pverkan p strldos och bildkvalitet av olika instllningar vid
kranskrlsrntgen och PCI. Utifrn dessa fynd snktes dosraten fr genomlys-
ning till en tredjedel. Dossnkningen utvrderades i en serie bestende av 154
PCI-procedurer fre och 138 efter optimeringen. Genom denna optimering
snktes det totala KAP-vrdet signifikant till tv tredjedelar av ursprungsvr-
det.
Sammanfattningsvis pvisar denna avhandling mjligheterna till dossnk-
ning i diagnostisk radiologi genom optimering av den radiografiska processen.

-8-
Preliminary reports

3 Preliminary reports
The following reports have been given:

Abstracts
1. Geijer H, Andersson T, Beckman K-W, Persliden J. En ny digital metod fr
scoliosrntgen - vrdering av strldoser. Hygiea Svenska Lkare-
sllskapets handlingar 1998;107(1):256-7. Swedish Medical Society,
Annual Meeting.
2. Geijer H, Andersson T, Beckman K-W, Persliden J. Strldosreducerande
tgrder vid angiografi. Hygiea Svenska Lkaresllskapets handlingar
1998;107(1):257. Swedish Medical Society, Annual Meeting.
3. Geijer H, Andersson T, Beckman K-W, Jonsson B, Persliden J. A new
digital method for scoliosis radiography - description and assessment of
radiation dose. Eur Radiol 1999;9 Suppl 1:S110. European Congress of
Radiology, Vienna, ECR 99.
4. Geijer H, Andersson T, Beckman K-W, Persliden J. Reduction of radiation
dose in angiography. Eur Radiol 1999;9 Suppl 1:S286. European
Congress of Radiology, Vienna, ECR 99.
5. Geijer H, Beckman K-W, Andersson T, Persliden J. Radiation dose and
image quality with a flat-panel amorphous silicon digital detector. Eur
Radiol 2001;11(2 Suppl 1):S280. European Congress of Radiology, Vienna,
ECR 2001.
6. Verdonck B, Nijlunsing R, Melman N, Geijer H. Image quality and X-ray
dose for translation reconstruction overview imaging of the spine, colon
and legs. In: Lemke HU, Vannier MW, Inamura K, Farman AG, Doi K,
editors. CARS 2001. Proceedings of Computed Assisted Radiology and
Surgery; 2001 June 27-30; Berlin, Germany. Amsterdam: Elsevier
Science; 2001. p. 500-5.

Posters
1. Geijer H, Beckman K-W, Jonsson B, Andersson T, Persliden J. Digital
scoliosis radiography: evaluation of a new method. Radiology
1999;213(P):500. Radiological Society of North America, Chicago, RSNA
99.

-9-
Hkan Geijer

4 Original papers
This thesis is based on the following original papers, which will be cited in the
text by their Roman numerals:

I. Geijer H, Beckman K-W, Jonsson B, Andersson T, Persliden J. Digital


radiography of scoliosis with a scanning method: Initial evaluation.
Radiology 2001;218:402-10.

II. Geijer H, Verdonck B, Beckman K-W, Andersson T, Persliden J. Digital


radiography of scoliosis with a scanning method: Radiation dose
optimization. Submitted.

III. Geijer H, Beckman K-W, Andersson T, Persliden J. Image quality vs


radiation dose for a flat-panel amorphous silicon detector: a phantom
study. Eur Radiol 2001;11:1704-9.

IV. Geijer H, Beckman K-W, Andersson T, Persliden J. Radiation dose


optimization in coronary angiography and percutaneous coronary
intervention (PCI). Part 1: experimental studies. Eur Radiol (in press).

V. Geijer H, Beckman K-W, Andersson T, Persliden J. Radiation dose


optimization in coronary angiography and percutaneous coronary
intervention (PCI). Part 2: clinical evaluation. Eur Radiol (in press).

- 10 -
Abbreviations

5 Abbreviations
ALARA as low as reasonably achievable
ANOVA analysis of variance
AP anteroposterior
DQE detective quantum efficiency
ESD entrance surface dose
ICS image criteria score
IQF image quality figure
KAP kerma area-product
keV kiloelectron volt
kV kilovolt (tube voltage)
LAO left anterior oblique
mA milliampere (tube current)
mAs milliampere second (tube charge)
MTF modulation transfer function
PA posteroanterior
PACS picture archiving and communicating system
PCI percutaneous coronary intervention
RAO right anterior oblique
RIS radiological information system
ROC receiver operating characteristics
ROI region of interest
SD standard deviation
TLD thermoluminescent dosimetry
VGA visual grading analysis

EU European Union
ICRP International Commission on Radiological Protection
ICRU International Commission on Radiation Units and Measurements

- 11 -
Hkan Geijer

- 12 -
Introduction

6 Introduction
6.1 General introduction
All radiological procedures involving X-rays deliver a radiation exposure to the
patient. In procedures where staff are present in the examination room, such
as in interventional radiological procedures, also the staff receive a radiation
exposure.
Radiation has long been known to be harmful to humans. The radiation
exposure received in an X-ray examination is known to increase the risk of
malignancy as well as, above a certain dose, the probability of skin damage and
cataract. The number of procedures performed is rising, and there is a trend
towards more complicated procedures that at the same time subject patients
and staff to higher doses. Some examples of these complicated, high-dose pro-
cedures are the wide range of interventional procedures such as percutaneous
coronary intervention (PCI), ablation procedures in the heart and vascular
interventions in other parts of the body such as aortic stent-grafting. There are
also a rising number of computed tomography (CT) procedures performed with
a concomitant rise in the collective dose. The CT issue is not investigated in
this document.
In todays diagnostic radiology there is a growing concern on radiation ex-
posure. This can be seen in the recommendations from the International
Commission on Radiological Protection (ICRP) where the current international
recommendations, ICRP 60, (ICRP 1991a) lay out the principles of justification
and optimization of all radiation exposures in diagnostic radiology. Later
recommendations, ICRP 85, (ICRP 2000) deal with the risk of skin damage in
interventional radiology. In Europe, the European Union council directive
97/43/Euratom (The Council of the European Union 1997) imposes the im-
portance of the two basic principles of justification and optimization.
The principle of justification implies that the benefits to the patient and
society of a radiological procedure must outweigh the risks for the patient as-
sociated with the radiation exposure. The other principle, the principle of opti-
mization, is the main topic of this document. This principle is introduced in
chapter 6.4 and is covered in detail later.
The present document presents investigations of the dose reduction possi-
bilities in three different areas. In papers I and II a digital method for scoliosis
radiography is described, and the radiation exposure to the patient is reduced
by changing exposure parameters. In paper III the potential in exchanging the
whole detector system for a recently available flat-panel detector is investi-
gated. In papers IV and V the potential for dose reduction in interventional
radiology is investigated using coronary intervention as a model. Altogether,
the whole project is focused on finding clinically useful solutions for optimi-
zation, leading to substantial dose reductions.
The project was undertaken at rebro University Hospital in the Depart-
ment of Radiology, which has been fully digital for nearly five years.

- 13 -
Hkan Geijer

6.2 Radiation dose


There are two main biological effects of ionising radiation: deterministic and
stochastic.
Deterministic effect: Cells in an organ or tissue are injured or killed by the
radiation, resulting in a loss of tissue function. A certain threshold dose has to
be exceeded before the effect is seen and above the threshold the severity of the
effect is dose-related. Skin damage and cataract formation are two types of
damage with a deterministic effect. This type of damage is almost exclusively
seen in interventional radiology, including interventional cardiology. Recent
years several reports have been published describing acute skin damage from
radiation (Shope 1996; Svik et al 1996; Va et al 1998; Wagner et al 1999,
2000; Koenig et al 2001a-b). The ICRP report 85 covers this issue in great de-
tail (ICRP 2000).
Stochastic effect: Instead of being killed, the irradiated cell may be modified
and continue to reproduce, potentially causing a malignancy. There is no
threshold level known, and the probability of the effect is linked to the radia-
tion dose. In other words, the risk for malignancy increases with increasing
dose, but the results of the effect is independent of dose the seriousness of
the malignancy is independent of the dose that caused it. If the initial damage
is to the germ cells in the gonads, hereditary effects may occur instead. This
issue is discussed in several ICRP reports (ICRP 1977, 1991a-b, 1996).
The risk of stochastic effect is decreased when the radiation exposure is
lowered. This also increases the safety margin to the production of determinis-
tic effects. However, certain aspects have to be considered when measuring
radiation dose depending on which effect is in focus. Also when lowering dose,
certain actions might influence one effect more than the other. These aspects
are covered in more detail later in this document.
Besides the exposure to the patient, there is also a radiation exposure to
staff. Mainly in interventional radiology, especially in cardiac procedures, the
radiation exposure to staff can be considerable. The source of this radiation to
staff is mainly scattered radiation from the patient. This subject is covered in
the chapters on coronary intervention.
In this document, the term radiation dose is sometimes used in a general
sense. It usually means a similar detriment as effective dose. When used in a
context, the dose quantity it represents is specified more in detail.

6.3 Image quality


In the diagnostic radiographic process there are several factors that must coin-
cide to reach a correct diagnosis: (1) a suitable examination must be chosen by
the radiologist in cooperation with the referring clinician, (2) image quality
should be adequate to show the pathology, (3) the observer must be able to see
and recognize this pathology, and (4) the result of the procedure should be
beneficial to the patient in some way. The image quality goal is thus just one
part in this long chain. This goal should furthermore be reached without sub-
jecting the patient to unnecessary radiation exposure.

- 14 -
Introduction

The second item above means that image quality must be quantified in
some way. This quantification can be done in several ways and might encom-
pass different parts of the imaging system or the entire system.
In digital radiographic systems, image quality is closely linked to radiation
exposure. Keeping other factors constant, image quality decreases with a lower
exposure because of the increased noise (quantum mottle) in the image.
Accordingly, image quality increases with an increased exposure. Digital X-ray
image quality is thus noise limited (Cowen 1991). In all attempts to reduce
dose, image quality is crucial. The dose reduction may not be pursued too far,
potentially jeopardizing the diagnostic outcome of the procedure. Conse-
quently, it is vital to follow the image quality while optimizing a radiographic
procedure. Unfortunately, methods for image quality assessment are less
straightforward than for dose measurement.

The image quality of X-ray systems can be characterized in several ways:


1. Physical characteristics such as contrast, spatial resolution (often ex-
pressed as modulation transfer function, MTF) and noise (expressed as
noise power spectrum, NPS, which is equivalent to the Wiener spectrum).
NPS is the noise variance analysed in terms of its spatial frequency con-
tent (ICRU 1996).
2. On a more system-wide approach, image quality is expressed as detective
quantum efficiency (DQE), which incorporates MTF, noise and exposure
level. DQE is regarded as a valuable tool to describe the performance of
digital radiographic systems (Moy 2000), but unfortunately it is difficult to
measure in clinical practice.
3. A more direct image-based method of evaluating overall system perform-
ance is by using contrast-detail phantoms (see chapter 8.2.2). These
phantoms contain test objects of different size and contrast, and the task
for the observer is to indicate the borderline visibility in a radiograph of
the phantom. These phantoms are commercially available and can be
used in a clinical context. A disadvantage (and, sometimes, an advantage)
is that they are evaluated by human observers with their inherent intra-
and inter-observer variability. This method has been used very frequently
in the present project. One way to avoid the human influence on the
readings is to use computerized reading of the images (Jansen and
Zoetelief 2000). Unfortunately, there are no commercially available com-
puter programs for this purpose.
4. The next step towards the clinical image is to use anthropomorphic phan-
toms. They lend themselves to subjective evaluation of image quality in an
optimization sequence. More objective methods that can be used with
anthropomorphic phantoms are Visual grading analysis (VGA, see chapter
8.2.4), Image criteria score (ICS) studies such as those following the Euro-
pean Union quality criteria (Commission of the European Communities
1996), and Receiver operating characteristics (ROC), which is a method
based on signal detection theory. An advantage of ROC studies is that
they measure the ability of observers to detect a signal in an image while
an obvious disadvantage is that they are very time-consuming. There are
reports stating that VGA can be used for evaluation of clinical image qual-
ity quite as well as ROC studies (Tingberg 2000).

- 15 -
Hkan Geijer

5. Images of real humans are the ultimate goal in finding objects for image
quality evaluation. They have the advantage of being the very same images
that are being used in clinical practice without any approximations. Dis-
advantages are the natural variability between patients and the increased
radiation dose if several settings are tested on one subject. The same
methods as for anthropomorphic phantoms, VGA, ICS and ROC, can be
used.

In the present project, image quality was assessed using a contrast-detail


phantom, an anthropomorphic phantom and with Visual grading analysis on
clinical images. These methods are discussed more in detail later.

6.4 The concept of optimization


The official publications mentioned above all stress the concept of optimization.
ICRP publication 26 states, The limitation of stochastic effects is achieved by
keeping all justifiable exposures as low as is reasonably achievable, economic
and social factors being taken into account (ICRP 1977). Similar words were
used in the European Union council directive 97/43/Euratom (The Council of
the European Union 1997) and they are known as the ALARA principle (As Low
As Reasonably Achievable). One interpretation of the ALARA principle, and the
one used in this document, is that the exposure to the patient should be
adjusted to obtain the required diagnostic information, not to get the best
image quality possible. The process of reaching this goal is called optimization
in this document.
Using traditional screen-film systems, the exposure is largely determined
by the need to get a sufficient blackening of the film; the detector dose must be
kept constant. It is possible to partly optimize a traditional screen-film system
by changing the quality of the X-ray beam or by changing the sensitivity of the
screen-film system. To do the latter on a larger scale, however, the whole
screen-film system has to be exchanged and in general, optimization is cum-
bersome using analog systems.
Optimization in a more thorough way, however, in fact requires digital
radiography (Martin et al 1999a). In digital radiography there is the possibility
of adjusting the exposure to a suitable image quality because of the wide dy-
namic range of the digital detectors; the detector dose may be varied. Instead,
image quality is more determined by the amount of noise in the image, assum-
ing unchanged radiation quality. If the detector dose is too low, increased noise
lowers image quality. A high detector dose usually results in better image qual-
ity but also a higher dose to the patient.
It has been shown that a high limiting resolution is not as necessary as
was believed earlier. Rather, it is more important to have a system that can
provide good image contrast over a wide latitude of exposures up to a modest
limiting resolution (Yaffe and Rowlands 1997). This fits in well with digital
radiographic systems, which in general dont have as high detail resolution as
their screen-film counterparts, but are much better in contrast resolution due
to the limited exposure latitude of film (Cowen 1991).
When installed at a location, the X-ray equipment is usually adjusted by
the manufacturer to standard factory settings to give a good image quality.

- 16 -
Introduction

This does not automatically mean the lowest possible radiation dose. Therefore,
an effort by the users to reduce the radiation dose to a level that results in a
diagnostically acceptable image quality without excessive radiation exposure is
often beneficial.
An alternative way of optimizing a radiological procedure is to change the
whole detector system. New systems with higher detective efficiency might lead
to lower doses without sacrificing image quality. This aspect is investigated in
the study on the flat-panel detector.
In interventional radiology the operator can influence the dose by the
amount of fluoroscopy and other imaging that is used. Consequently, the
operator's effect on dose can also be optimized. Normally this is done through
education and training. This potential is investigated in the study on coronary
intervention.

6.5 Scoliosis radiography


Scoliosis radiography is mainly used for measurement of Cobb angles in the
spine. These angles are used as an aid in setting the diagnosis and as follow-
up during the treatment process with braces or surgery.
Earlier, scoliosis radiography was generally performed using a screen-film
system with a large cassette. In a digital environment with use of storage
phosphor image plates, scoliosis radiography is difficult to perform because
there are no detector plates large enough to cover the entire spine, with the
exception of the spine of small children. Scoliosis radiography has been
performed also by using image intensifiers but with the same type of size
restrictions. Previously described techniques to overcome this problem usually
have been performed with dedicated equipment (Kushner et al 1986) or by
combining several storage phosphor plates (Dewaele et al 1999).
When performing scoliosis radiography, the ALARA principle for a low dose
is especially important since the patients are often subjected to repeated
studies during a treatment period. In the young age when scoliosis is typically
evolving, a low radiation dose is even more important due to the greater
radiation sensitivity and higher risk for developing fatal cancer at young age
(BEIR 1990; ICRP 1991b). Several studies have attempted to reduce the dose in
scoliosis radiography, mainly by changing the sensitivity of a screen-film
system. Because of the inherent larger dynamic range in digital radiographic
systems, a dose reduction is easier to achieve (Kogutt et al 1989; Kalmar et al
1994; Stringer et al 1994). Furthermore, scoliosis radiography offers good
possibilities for radiation dose reduction, since the images are mostly used for
angle measurement instead of detailed anatomical diagnosis.
One method to overcome the problem of limited detector size is the method
investigated in papers I and II. It was first described by van Eeuwijk et al (van
Eeuwijk et al 1997) and consists of a multiple-image scan of the entire spine
using digital image intensifier-based general radiographic equipment. The
images are then reconstructed into one image at a workstation where the angle
measurements can be made.

- 17 -
Hkan Geijer

6.6 Flat-panel detector


Digital radiography in the form of storage phosphor plates opens up a
possibility for dose reduction through a wider dynamic range compared to
screen-film systems. Still, the detective quantum efficiency (DQE) of storage
phosphor plates is not much better than screen-film (Neitzel et al 2000). With
the recent introduction of flat-panel direct digital X-ray detectors, however,
there is a real possibility of both dose reduction and improved image quality,
because these flat-panel detectors have a higher DQE, both compared to
screen-film and storage phosphor plates (Neitzel et al 2000; Spahn et al 2000).
Dose reductions of over 50% have been predicted with the flat-panel detector.
Paper III describes a study of the radiation dose image quality relationship for
the flat-panel detector compared to storage phosphor plates and a screen-film
system. The study is clinically oriented and was performed at a clinical
installation of a flat-panel detector.

6.7 Coronary intervention


Coronary angiography and percutaneous coronary intervention (PCI) are
procedures with a high radiation dose, both to the operator and patient. The
radiation risks associated with these procedures are the stochastic risk of
cancer induction, both to the patient and the operator, radiation skin injuries
in the patient, and cataract development in the operator. The increased use of
radiography and fluoroscopy in interventional procedures further increases
these problems. There is a growing concern about the high radiation doses in
interventional radiology since they can be contrary to the ALARA principle
(ICRP 2000; Wilde et al 2001). Furthermore, severe skin reactions from
interventional fluoroscopy have been reported (Koenig et al 2001a-b) indicating
the problem with high doses. Modern angiographic equipment has a high
radiation output capacity, which increases the potential for radiation damage
but also makes some radiation protection methods possible.
In angiography, the radiation dose is to a great extent depending on the
behaviour of the operator, who can influence the amount and quality of
fluoroscopy. PCI is an extreme example of this, since more than half of the
radiation dose originates from fluoroscopy (Cusma et al 1999). Thus, radiation
dose in PCI is depending both on technical factors and operator behaviour.
In the present study, an evaluation of radiation dose and image quality
was made for different parameters in a simulated coronary catheterisation. Ac-
cording to these findings, the examination situation in coronary angiography
and PCI was optimized, both regarding technical parameters and operator be-
haviour. The resulting dose reduction was evaluated in a series of clinical PCI
procedures.

- 18 -
Aims

7 Aims
The overall aim of this thesis is to find ways to optimize the radiation dose -
image quality relationship in diagnostic radiology. All solutions are explicitly
meant to be useful in daily clinical practice. The methodology for optimization
is evaluated in various clinical situations:

In scoliosis radiography, where the patients are young and thus extra
sensitive to radiation, the optimization is performed through adjustment of
the equipment.
In the second main part, dealing with the flat-panel detector, the optimiza-
tion potential in exchanging a whole detector system is evaluated. Such an
exchange has implications for a wide range of radiographic procedures.
In coronary intervention, where there is a risk of both acute skin burns
and radiation-induced malignancy, the possibility of optimizing an inter-
ventional procedure is investigated.

A second aim is to find methods for dose measurement and image quality
assessment that are practical to use during the optimization process.

- 19 -
Hkan Geijer

8 Materials and Methods


In the first two parts of this chapter, the methods for dose measurements and
image quality assessment are described. In the third, fourth and fifth parts, the
methods for the three main parts of the project (scoliosis radiography, flat-
panel detector, and coronary intervention) are described. Finally, the statistical
methods and ethical considerations are covered.

8.1 Radiation dose


The measurement of radiation dose in various forms is a fundamental part of
this project, and a recent review of dosimetry methods is also available
(Faulkner et al 1999). The following methods were the ones normally in use.

8.1.1 Entrance dose and ESD


The entrance dose, also known as the incident dose, is the air kerma (kinetic
energy released per mass unit) at the focus to skin distance without the patient
present (ICRP 2000). It does not include back-scattered radiation from the
patient.
Entrance surface dose (ESD) is the absorbed dose in the skin at a given
location on the patient. It includes back-scattered radiation from the patient.
Entrance dose can be converted to ESD by multiplying with a backscatter fac-
tor and converting from air to tissue. The backscatter factor depends on beam
quality and field size and it can be found in tables (IPSM 1988).
ESD and entrance dose are point quantities and they are thus independent
of collimation, apart from the influence on the backscatter factor. ESD is re-
garded as the best indicator of deterministic effects such as skin burns (ICRP
2000). Both entrance dose and ESD decrease when the focus-to skin distance
is increased due to the inverse square law. The unit of both is the gray (Gy).

In the project, these parameters were measured in three ways:


Entrance dose was measured using a PMX-III instrument with a solid-
state detector (RTI Electronics, Mlndal, Sweden) which was put at the
entrance surface of the patient or phantom. The dose value with the PMX
instrument is entrance dose without backscatter from the patient or phan-
tom. The instrument was calibrated with a calibration traceable to a stan-
dard laboratory.
ESD was measured using thermoluminescent dosimeters (TLD). The TLD
discs are about 5 mm in size and can easily be placed on the surface of a
phantom or a patient. Normally, they cannot be seen in the radiographic
image and thus do not disturb a clinical procedure. The output from the
TLD discs is then measured in a TLD reader. Because the TLD discs are
sensitive in all directions, the dose value from a TLD measurement in-
cludes backscatter.
In paper IV, ESD was calculated from the kerma area-product (KAP) by di-
viding with the field size and adding the backscatter component.

- 20 -
Materials and Methods

8.1.2 Absorbed dose from scattered radiation


The scattered radiation from a phantom or patient was measured as air kerma
using an ionisation chamber (Series 1015C X-ray Monitor with a 10X5-180
ionisation chamber, Dosimeter Corporation of America, Cincinnati, OH, USA)
(Figure 1). The ionisation chamber was placed in the position of the operator
and all measurements were performed free in air without backscatter.

Figure 1. The instrument that was used for


measurement of scattered radiation. The ioni-
sation chamber is shown on a tripod. An
Alderson phantom can be seen behind the pro-
tective screens used in coronary intervention.

8.1.3 Kerma area-product


Kerma area-product (KAP) is the product of the entrance dose and the area of
the X-ray field (ICRP 2000). The unit of KAP is Gy cm2. In fluoroscopy, KAP
rates can be recorded in which case the unit is Gy cm2/s. A similar quantity is
dose area-product (DAP). In the energy range of diagnostic radiology, the two
values are numerically equivalent.
KAP is normally measured with an ionisation chamber placed in the pri-
mary X-ray beam. Because KAP is a measurement of an area dose, it is very
sensitive to changes in field size such as collimation. On the other hand it is
not affected by variations in focus-to skin distances. KAP measurements are
easy to obtain since the ionisation chamber can be attached to the collimator
housing and no probes have to be put onto the patient. KAP measurements are
very valuable in a clinical study as an indicator of stochastic effects such as
induction of malignancy (ICRP 2000).
In the project, KAP values were recorded in several studies. Mostly, KAP
was measured using a transmission ionisation chamber placed close to the X-
ray tube. It was connected to an electrometer, Doseguard 100 (RTI Electronics,
Mlndal, Sweden) (Figure 2). The instrument was calibrated with a calibration
traceable to a standard laboratory.

- 21 -
Hkan Geijer

In the study for paper V, a KAP meter built into the collimator housing of
the angiographic equipment was used (Diamentor, PTW, Freiburg, Germany).
KAP values, separated into fluoroscopy and acquisition doses, can be printed
together with other details of the examination. The instrument was calibrated
with a calibration traceable to a standard laboratory.

a b

Figure 2. The instrument that was used for measurement of KAP. (a) Electrometer. (b) Ionisation
chamber mounted on the collimator housing.

8.1.4 Effective dose


The effective dose is regarded as the best indicator of stochastic risk such as
induction of malignancy. It is defined as the sum of the weighted equivalent
doses in all tissues and organs of the body. The calculation of effective dose is
described in detail in ICRP report 60 (ICRP 1991a) with extensions in ICRP
report 73 (ICRP 1996).
The basic physical quantity used in radiological protection is the absorbed
dose, DT, averaged over an organ or defined tissue, T, where DT is the energy
deposited in the organ divided by the mass of that organ. The unit of absorbed
dose is called the gray (Gy).
Some radiations are more effective than others in causing stochastic ef-
fects. To allow for this, a further quantity has been introduced. This is the
equivalent dose, HT, which is the average absorbed dose in an organ or tissue
multiplied by a dimensionless radiation weighting factor, wR. For photons, the
radiation weighting factor is 1. The unit of equivalent dose is called the sievert
(Sv).
Radiation exposure of the different organs and tissues in the body results
in different probabilities of harm. To reflect this, the equivalent dose in each
organ and tissue is multiplied by a tissue weighting factor, wT (Table 1), and
the results are summed over the whole body to give the effective dose, E. It is
given by the expression

E = wT H T (1)
T

- 22 -
Materials and Methods

The unit of effective dose is called the sievert (Sv).


Consider a partial body exposure to irradiation in which only certain or-
gans were irradiated. The effective dose is a calculated quantity, which if deliv-
ered uniformly to the whole body would give the same hypothetical risk as the
partial body exposure.

Tissue or organ Tissue weighting Table 1. Tissues, organs and tissue weighting factors
factor, wT for calculation of effective dose.
a The remainder is comprised of the following addi-
Gonads 0.20
Bone marrow (red) 0.12 tional tissues and organs: adrenals, brain, upper
large intestine, small intestine, kidney, muscle, pan-
Colon 0.12
creas, spleen, thymus and uterus.
Lung 0.12
Stomach 0.12
Bladder 0.05
Breasts 0.05
Liver 0.05
Oesophagus 0.05
Thyroid 0.05
Skin 0.01
Bone surface 0.01
a
Remainder 0.05

8.1.5 Alderson phantom


In order to calculate the effective dose, absorbed dose must be measured in the
relevant organs according to Table 1. This can be done using an anthropomor-
phic phantom, which is composed of materials simulating the attenuating
properties of the human body. Normally, it is sliced to facilitate the insertion of
TLD discs.
In papers I and IV, absorbed dose was measured using an Alderson phan-
tom (Alderson et al 1962; ICRU 1992). The phantom was loaded with TLD discs
(TLD-100, Harshaw Chemical Co., Solon, OH, USA) in the relevant organs to
measure absorbed dose. The TLDs were individually calibrated and read with a
TLD reader (Alnor Dosacus, Alnor Oy, Turku, Finland or Harshaw 5500,
Harshaw/Bicron, Solon, OH, USA). The effective dose was then calculated ac-
cording to equation 1.
Two types of Alderson phantoms were used. In paper I, a phantom emu-
lating a female of height 163 cm and weight 54 kg was used (Figure 3). In
paper IV, a phantom emulating a male of height 175 cm and weight 73.5 kg
was used.
Energy dependence: The Alderson phantoms were developed for
measurements in radiotherapy with higher photon energies (Alderson et al
1962). The attenuation properties in the diagnostic energy range differs
somewhat from human tissue. For the heavily filtered X-ray beam used in
these studies, this difference is small with water having a linear attenuation
coefficient that is only 1% lower (Shrimpton et al 1981).

- 23 -
Hkan Geijer

Figure 3. The female Alder-


son phantom that was used
for measurement of ab-
sorbed dose. Here it is
shown in a set-up for
scoliosis radiography.

8.1.6 Monte Carlo simulation


The Monte Carlo method in this context is a simulation of the passage of indi-
vidual X-ray photons through the body (Persliden 1986; Andreo 1991). A com-
puter program is used to follow the individual photons along their passage
through, or absorption in the body. The fate of each photon is randomised ac-
cording to the probability of different effects: scattering, absorption or trans-
mission. The interaction probability depends on the photon energy and inter-
acting material. By simulation of a large number of photon histories, eg.
50,000 photons in an X-ray field, a mean value of the absorbed energy in a
specific organ and the effective dose can be calculated.
In the present project, organ doses and effective dose were calculated with
the PCXMC computer program (Finnish Radiation and Nuclear Safety Author-
ity, Helsinki, Finland) (Tapiovaara et al 1997; Anonymous 1999). As input val-
ues, ESD obtained with TLD measurements were normally used. The values
were corrected to get entrance dose, which is required by the program.

8.1.7 Equivalent diameter and Reference Man


Dose data from a patient cohort can be heterogeneous because of differences in
height and weight. One method to reduce the variability that depends on
patient size is to transform the height and weight data of patients into
equivalent diameter De (Lindskoug 1992) using the formula

M 1000
De = 2 (2)
H

where De = equivalent diameter in cm, M = body mass in kg, and H = body


height in cm.

- 24 -
Materials and Methods

The resulting equivalent diameter can be compared to a standard entity


named Reference Man (ICRP 1974), which has a De of 22.9 cm (70 kg, 170 cm).
The dose values can be transformed to the Reference Man dose, that is the
dose that the patient would have received had he or she been of the Reference
Man size, using the formula

kD
e e ( ref )
KAPref = KAPmeas kDe ( meas) (3)
e
where KAPref = KAP for Reference Man size, KAPmeas = measured KAP, k =
equipment-specific constant, De(ref) = equivalent diameter of Reference Man
(22.9 cm), and De(meas) = equivalent diameter of the patient (Chapple et al 1995).
The equipment-specific constant has to be determined for each type of
equipment.

8.2 Image quality


The measurement of image quality has been reviewed recently (Martin et al
1999b). Following are the methods for image quality evaluation that were used
in the present project.

8.2.1 Image viewing conditions


All digital images were evaluated at a workstation connected to a Picture
Archiving and Communicating System (PACS) from Sectra/Philips Medical Sys-
tems, Linkping, Sweden. The observers were allowed to manipulate the con-
trast/greyscale and zoom the images in the PACS workstation. Images were
viewed on a 21-inch 1,280 x 1,024 pixel monitor with 8 bits greyscale. In the
project, three different brands of monitors were used:
A greyscale monitor from Image Systems, Hopkins, MN, USA
Dell Ultrascan P1110, Austin, TX, USA, a colour monitor.
Sony CPD-E500E, Tokyo, Japan, a colour monitor.
The screen-film images were viewed at a light-box where masking could be
used freely.

8.2.2 Contrast-detail phantoms


Contrast-detail phantoms are used to evaluate two very important parameters
of image quality: contrast and detail resolution. This type of phantom usually
consists of an acrylic sheet with holes drilled in various configurations. There
are several different phantoms commercially available such as the Leeds
phantom and the CDRAD phantom. In this project only the CDRAD phantom
was used.

- 25 -
Hkan Geijer

Figure 4. Radiograph of the con-


trast-detail phantom that was
used for image quality assess-
ment.

In all studies, part of the image quality assessment was done with a
CDRAD 2.0 contrast-detail phantom (Instrumentele Dienst, Nijmegen, the
Netherlands) (Thijssen et al 1989). The phantom consists of a 265 mm by 265
mm by 10 mm acrylic sheet with drilled holes of different depth and diameter.
In each column, the diameter of the drilled holes varies logarithmically from 8
to 0.3 mm. In each row, the depth varies logarithmically from 8 to 0.3 mm. A
radiograph of the CDRAD phantom is shown in Figure 4. In each square, ex-
cept in the top three rows, there is one hole in the middle and one hole ran-
domly located in one corner. The observer has to indicate in which corner the
hole is located, thus making a four-alternative forced choice (4-AFC) (Aufrichtig
1999). All detection results were corrected according to the user manual for the
CDRAD phantom. Each correctly detected target needs two or more correctly
detected nearest neighbours (out of four) to remain true, and each incorrectly
or not detected target is considered true if it has three or four correctly
detected nearest neighbours. In this way, small inconsistencies in the readings
are eliminated.
From the resulting data, a graph can be drawn of the just visible objects,
see Figure 12 on page 41 for an example. Furthermore, a numerical value, the
Image Quality Figure (IQF), can be calculated (Thijssen et al 1998). The IQF is
defined as

15
IQF = Ci Di ,th (4)
i =1

where IQF = Image Quality Figure, i = contrast-column number, Ci = contrast


(depth of hole), and Di,th = threshold diameter in contrast-column i.

- 26 -
Materials and Methods

Figure 5. Radiograph of an anthropo-


morphic phantom that was used for
image quality evaluation.

Because the IQF is derived from the sum of the products of depth and
diameter for the just visible objects in the phantom, a lower IQF indicates bet-
ter image quality.
The phantom was normally placed between two 7.5 cm-thick acrylic sheets
to simulate the scattering conditions of the human body. Normally, three im-
ages were produced at each setting to reduce the effect of random photon
fluctuations. Also to reduce random variations, each image was read by several
observers. The IQF results for all observers were then averaged.
Parallax effect: Since the smallest and deepest holes of the CDRAD phan-
tom are located at the lower right corner, the farthest away from the centre
beam, there is a potential for parallax effect since the divergent beam would
strike the holes at an angle (Neitzel et al 2000). This potential was investigated
in paper III by exposing images centred in different parts of the phantom.

8.2.3 Anthropomorphic phantoms


Anthropomorphic dose measurement phantoms such as the Alderson phan-
toms were created for dose measurement purposes in radiotherapy. Because
they consist of 2.5 cm thick slices, there are disturbing artefacts in the images,
making image quality assessment difficult. However, anthropomorphic phan-
toms exist that are not sliced. These phantoms can be used as an easy means
of comparing image quality between different equipment settings although nu-
merical results cannot be produced directly. This type of comparison was made
informally in the studies for papers I, II and III with an Alderson Phantom
Patient. A sample image using the flat-panel detector is shown in Figure 5.

- 27 -
Hkan Geijer

8.2.4 Visual Grading Analysis (VGA)


Image quality in clinical scoliosis radiography images was evaluated using Vis-
ual Grading Analysis (VGA) (Manninen et al 1984; Tingberg 2000). VGA is a
powerful method of comparing image quality between two images. One of these
images is treated as a reference image and the other image is scored according
to a classification system. Ideally, the test image and the reference image
should be images from the same patient but with different radiographic set-
tings, if the goal is to compare settings. This was the case in one part of paper I
where ten patients were imaged the same day using two different protocols,
both digital exposure and digital fluoroscopy. The disadvantage of this
approach is the increased radiation dose to the patient.
For the rest of the VGA studies, one image was chosen as the reference
image. This image was selected to be an image of subjectively good quality.
Then all images in the image base were compared to this reference image.
Using this method it was possible to compare the results for two different pro-
tocols even if no patients were imaged with both protocols. This method was
used in paper I when comparing digital exposure to screen-film and in paper II
comparing the baseline digital exposure protocol to the optimized protocol.
In VGA, the observer is often forced to grade certain parts of the image.
This can be accomplished through masking of the image. Options differ
whether this is beneficial or not. In the present study, no masking was used.
Image quality is normally graded by visibility of anatomical structures or
the visual quality of certain aspects, eg. noise or contrast. The grading can be
annotated using an arbitrary scale such as 2, -1, 0 +1, +2 or 1 to 5 for a five
grade scale. See chapter 10.4 for comments on the statistical treatment of
these results.

8.3 Scoliosis radiography


The study in paper I was designed to evaluate the radiation dose, image quality
and accuracy in angle measurements for three different protocols for scoliosis
radiography, two digital and one screen-film method. The digital protocols are
hereafter called digital exposure and digital fluoroscopy. The screen-film
method was used as a reference in the comparison with the digital methods. In
paper II, the digital exposure protocol was optimized to a lower radiation dose
with still acceptable image quality. The resulting new protocol is hereafter
called the optimized protocol.

8.3.1 Radiographic techniques


All digital imaging was performed using a Multi Diagnost 4 (Philips Medical
Systems, Best, the Netherlands) digital image intensifier-based unit with capa-
bilities for exposure and grid-controlled pulsed fluoroscopy (GCF). The patient
stands erect facing the image intensifier in a posteroanterior (PA) projection
(Figure 3 on page 24). Because of the undercouch tube set-up of this
equipment, the X-ray beam passes the tabletop before entering the patient. The
whole spine is exposed with a series of images synchronised with a cranio-
caudal scan of the X-ray tube and image intensifier C-arm. This sequence
produces 30-60 images depending on image frequency and scan speed.

- 28 -
Materials and Methods

The main difference between digital exposure and fluoroscopy is that the
fluoroscopy protocol employs grid-controlled pulsed fluoroscopy with a much
lower dose request. The grid-control feature automatically modulates exposure
parameters during the pulse. Each fluoroscopic image is saved to memory.
All images are then transferred to a workstation (EasyVision, Philips Medi-
cal Systems) where they are reconstructed to an overview image with the aid of
pattern recognition (Verdonck et al 2001) (Figure 6).
For comparison purposes, a screen-film examination was also evaluated. It
was performed in PA projection with a large dedicated film cassette giving a
radiated field of about 25x70 cm.
Images acquired with the three methods are shown in Figure 7.

8.3.2 Dose and image quality assessment


Several dose values were measured:
KAP values were measured with a Doseguard instrument.
Effective dose was calculated from absorbed dose measured with a female
Alderson-type phantom, loaded with TLD discs in relevant organs. This
measurement was not performed for the optimized protocol in paper II
since a very good correlation between measured values and Monte Carlo
simulations had been demonstrated in paper I.
TLDs were also placed on the dorsal aspect of the phantom to measure
ESD.
Entrance dose data were used as input values for Monte Carlo simula-
tions. One Monte Carlo simulation had to be performed for each image.

Figure 6. Reconstruction principles


for scoliosis radiography. The im-
age on the left shows the individual
images that were merged to form
the image on the right.

- 29 -
Hkan Geijer

Figure 7. Images acquired with the


different methods. (a) Screen-film.
(b) Digital exposure. (c) The same
patient, imaged the same day with
digital fluoroscopy.

a b c

a b
Figure 8. Photograph (a) and radiographic image (b) of the phantom used for the angle meas-
urements. The phantom is composed of acrylic, with aluminium sheets simulating vertebrae.

- 30 -
Materials and Methods

Image quality was evaluated both with the CDRAD contrast-detail phan-
tom, where one image of each setting was read by three independent observers,
and using Visual grading analysis (VGA) in clinical images. An image base was
formed from 30 screen-film images and 30 images acquired with the digital ex-
posure protocol. These images were all compared to a reference screen-film im-
age of good quality. This approach was chosen because it was not possible to
examine patients with both screen-film and a digital method. Ten patients were
after informed consent double examined with both the digital exposure and the
digital fluoroscopy protocol. This permitted direct comparison of image quality
within each patient, with the digital exposure image being the reference image
in each pair. In paper II, the image base consisted of 30 images acquired with
the digital exposure protocol (same as in paper I) and 30 images with the opti-
mized protocol. These were all compared to a digital exposure reference image.
In both papers, each image was evaluated separately and independently by
three experienced radiologists.
Different approaches were employed in papers I and II as to which parts of
the image should be evaluated. In paper I, three levels of the spine were evalu-
ated in each image; third thoracic vertebra (T3), tenth thoracic vertebra (T10),
and third lumbar vertebra (L3). When evaluating the results it was found that
the results for T10 and L3 were very similar. The image quality at level T3 was
much inferior for screen-film because of a much darker exposure, but the
comparison with film was not an issue in paper II. Therefore it was decided to
allow the use of the whole image for the VGA study in paper II: the observers
were allowed to choose which vertebra to compare as long as the same part
was evaluated in the reference image. The motivation for this was that the pur-
pose of the study was to evaluate image quality and not positioning errors or
differences due to malformations or different degree of scoliosis. No masking
was used.
In each level, image quality was evaluated as six different aspects; contrast
level, noise, sharpness, visualisation of vertebral end plates, subjective suit-
ability for judging skeletal abnormalities and subjective suitability for angle
measurements. Furthermore, the overall quality of the image was assessed.
Each aspect was independently scored on a five level scale as much worse,
worse, equal, better or much better compared with the reference image
quality.

8.3.3 Angle measurements


The accuracy of Cobb angle measurements was evaluated in paper I with a
phantom made from 13.5 cm of acrylic with two-mm thick aluminium sheets in
the middle simulating two vertebrae (Figure 8). Thirty different angles between
0 and 53 were randomly set and imaged with all three methods. Each angle
was independently measured once by three observers.
The clinical images used for VGA in paper I were also used for angle
measurements. Each angle was measured twice by three observers.

- 31 -
Hkan Geijer

8.3.4 Optimization of the digital exposure protocol


The digital exposure protocol was optimized in paper II in the following way.
1. Lowered dose request: The dose request was lowered from about 50% to
33% of the normal dose setting. This would in theory result in a dose re-
duction of 33%.
2. Increased pulse frequency and scan speed: In order to reduce the amount
of overlap between consecutive images, the pulse frequency was increased
from 2 to 3 images/s while the scan speed was increased from nominally 4
to 7.5 cm/s. One effect of this change might be increased parallax arte-
facts. Isolated, this change would in theory result in a dose reduction of
30%.
3. Fixed mA: The tube current was changed from falling load to fixed 160 mA
to avoid too high currents leading to too short exposure times. For optimal
imaging, the exposure time should be between about 5 and 15 ms. In the-
ory, image quality should improve but no dose reduction is to be expected.
4. Higher kV: A higher tube voltage increases X-ray penetration and reduces
dose efficiently but at the cost of lower image contrast. The voltage was in-
creased from 70 to 81 kV. In theory, this would result in a 14% reduction
of the effective dose as calculated with a Monte Carlo simulation tech-
nique.

8.4 Flat-panel detector


In the study for paper III the dose image quality relationship for a flat-panel
detector was evaluated using a contrast-detail phantom. For comparison, stor-
age phosphor plates and screen-film were also evaluated.

8.4.1 Radiographic techniques


The evaluated system is a flat-panel detector (Chaussat et al 1998; Spahn et al
2000) (Pixium 4600, Trixell, Moirans, France) mounted in a bucky table (Digital
Diagnost, Philips Medical Systems). The flat-panel detector is a clinical installa-
tion integrated in a PACS. Image quality and radiation dose was compared to
ST-V storage phosphor plates (Fuji Photo Film Co., Tokyo, Japan) in a PCR sys-
tem (AC-3, Philips Medical Systems) and a screen-film system (Kodak, Roches-
ter, NY, USA, speed 160).
The flat-panel detector has a size of 43 by 43 cm with a matrix of about
3,000 by 3,000 pixels, which gives a pixel size of 143 m.
To simulate the scattering conditions of a human body, 16 cm of acrylic
was used for all imaging. For the flat-panel detector and storage phosphor
plates, the standard setting for intravenous urography with automatic expo-
sure control was used. The flat-panel detector could be set at exposure levels
equivalent to a 400-, 600- or 800-speed system. All three settings were evalu-
ated. With storage phosphor plates, the default sensitivity setting, about 200,
was used. With the screen-film system, the film was exposed manually in the
bucky table to get a sufficient blackening, which resulted in an optical density
of 1.9. Three images were exposed at each setting to compensate for random
fluctuations in noise.

- 32 -
Materials and Methods

8.4.2 Dose and image quality assessment


Entrance dose for the acrylic slab was measured with a PMX-III instrument.
Image quality was assessed with a CDRAD 2.0 contrast-detail phantom.
Four observers (two radiologists and two medical physicists) read all images
independently giving 12 scores for each setting (three images and four observ-
ers). All digital images were evaluated at a PACS workstation using one of two
monitors: observer 1 and 3 used an Image Systems monitor while observer 2
and 4 used the same monitor, a Dell colour monitor.
The effect of parallax was investigated by exposing images using the flat-
panel detector at speed 400. Three images were centred halfway between the
centre and the lower right corner of the CDRAD phantom, three images were
centred at the lower right corner, and three images were exposed at a focus-
detector distance of 100 cm instead of the standard 110 cm. All images were
read independently by four observers, the same as above.

8.5 Coronary intervention


The study for papers IV and V consisted of an optimization of the dose image
quality relationship in cardiac catheterisation and percutaneous coronary
intervention (PCI). The approach was to reduce the dose rate for fluoroscopy
without changing the settings for digital acquisition.

The study consisted of three main phases:


Phase 1: Prior to optimization
Measurement of primary and scattered radiation with different equipment
settings
Calculation of effective dose in a simulated PCI procedure
Monte Carlo simulation
Evaluation of image quality for different fluoroscopy settings
Recording of KAP values in a clinical population
Phase 2: Optimization
Adjustment of dose settings for fluoroscopy
Phase 3: After optimization
Calculation of effective dose in a simulated PCI procedure
Monte Carlo simulation
Recording of KAP values in a clinical population

8.5.1 Radiographic techniques


The study was performed using a digital cardiac imaging system (HM-3000,
Philips Medical Systems). All imaging is digital with digital acquisition
performed at 12.5 images/s. Images are eventually stored on a Compact Disc.
The operator can choose between three different fluoroscopy modes when
performing fluoroscopy. These modes include continuous and pulsed
fluoroscopy as well as the possibility of introducing filters containing
aluminium and copper into the beam during fluoroscopy. All digital acquisition
is performed without additional filtration.

- 33 -
Hkan Geijer

Figure 9. To the left of the


author the two screens used
during cardiac catheterisation
can be seen. They are covered
in sterile drapings.

The examination table is fitted with a lead curtain (Scanflex/Borntech,


Stockholm, Sweden) with a lead equivalence of 0.5 mm in the upper part and
0.3 mm in the lower part. A ceiling-mounted arm supports a lead acrylic screen
(Mavig, Munich, Germany) with a lead equivalence of 0.5 mm. The set-up of
these screens is shown in Figure 1 and Figure 9.

8.5.2 Dose measurements


Primary and scattered radiation with different equipment settings: An Alderson
phantom was used to simulate the scattering conditions of a human patient.
The phantom was irradiated at different equipment settings that might be used
in a typical coronary angiography or PCI.
The patient dose rate was measured as KAP rate using a Doseguard
instrument.
The absorbed dose rate for the operator (due to scattered radiation) was
measured with an ionisation chamber. The ionisation chamber was placed
at chest level in the typical position of the operator close to the groin of the
patient (Figure 1 on page 21).
ESD rate was calculated from the KAP rate with correction for the
entrance field size.
Entrance dose rates were also used for Monte Carlo simulations of the
rate of effective dose for various settings.

Influence on KAP rate, scattered radiation and in some cases ESD rate and the
rate of effective dose was evaluated for the following settings:
Various projections that are normally used in cardiac catheterisation and
PCI.
With or without protective screens between patient and operator.

- 34 -
Materials and Methods

a b

Figure 10. Sample images from the PCI procedures with the Alderson phantom for calculation of
effective dose. (a) Before optimization. (b) After optimization. A collimation of about 7 mm at
each side has been introduced.

Filtration in fluoroscopy with an addition of 1.5 mm Al + 0.1 mm Cu or


1.5 mm Al + 0.4 mm Cu.
Image intensifier field size.
Collimation, 2 cm on each side.

Calculation of effective dose in a simulated PCI procedure: The effective dose for
a standard PCI procedure before and after optimization was calculated from
absorbed dose measured in relevant organs in the Alderson phantom as
outlined in 8.1.4 and 8.1.5. TLDs were also placed in the entrance fields
dorsally to record ESD.
The median KAP for a standard PCI procedure was obtained from a series
of 220 PCI procedures in our institution. Of these, 74 procedures included a
coronary angiography as part of the procedure. Before optimization, all settings
were those set at installation of the equipment and the aim was to irradiate the
phantom to obtain the median KAP values for fluoroscopy and digital
acquisition from the clinical series (Table 7 on page 47). The KAP values were
divided equally in four different projections that are used in PCI. After
optimization of equipment settings, the phantom was irradiated again with the
new settings. This time, the aim was to irradiate with the same fluoroscopy
time and the same number of images in each projection as before optimization.
In order to simulate an effect of improved user behaviour, a slight collimation
of about 7 mm at each side was introduced. This amount of collimation was
chosen as a clinically reasonable collimation that could be used in virtually all
PCI procedures. Sample images are shown in Figure 10.
Conversion factors from KAP to effective dose were calculated for the two
situations before and after optimization. These factors were applied to the KAP
values from the clinical series to estimate the effective dose.

- 35 -
Hkan Geijer

Scattered radiation to the operator was also measured in the same session.
In the session prior to optimization, no protective screens were used in order to
create a worst case scenario, which was in fact standard practice in our
institution until a few years ago. In the session after optimization, screens were
used just like in an ordinary PCI procedure in our institution.
Monte Carlo simulation: Effective dose for a typical PCI procedure was also
simulated with Monte Carlo methodology before and after optimization. ESD
was measured with TLDs using the Alderson phantom in the same session as
for measurement of absorbed dose. Each of the four projections was measured
separately and since entrance doses had to be separated in fractions for
fluoroscopy and digital acquisition in each projection as a basis for the Monte
Carlo calculation, these fractions were assumed to be proportional to the KAP
values for fluoroscopy and acquisition. Equivalent doses and effective dose
were calculated with the PCXMC computer program. All Monte Carlo
calculations were tailored to a male patient of the same size as the Alderson
phantom.

8.5.3 Image quality assessment


For evaluation of image quality, the CDRAD contrast-detail phantom was used.
It was placed in the middle of a 15 cm thick slab of acrylic to simulate the
scattering conditions of the human body. One setting, however, was also read
with 20 cm of acrylic to evaluate the effect of a larger patient. Three observers
(two radiologists and one medical physicist) viewed the phantom during fluoro-
scopy at 18 cm image intensifier size while recording KAP rates.

8.5.4 Optimization of settings


The following settings were changed in order to lower dose:
The equipment was set always to start at the lowest of the three selectable
fluoroscopy dose rates (the button labelled Low) with the option of
manually increasing dose in two steps. Before optimization the equipment
had started at the middle button (labelled Normal) with the option of
increasing or decreasing the dose rate manually.
Pulsed fluoroscopy was used for the lowest setting since image quality
evaluation had shown a good balance between dose and image quality.
The extra filtration for fluoroscopy was increased from 1.5 mm Al + 0.1
mm Cu to 1.5 mm Al + 0.4 mm Cu for the start-up dose setting.
For the lowest dose setting, the image intensifier dose request was
reduced to half the original value.
These changes are shown in Figure 17 on page 48 as the change from setting
NP to setting LP-R. The optimized settings are now used as standard settings in
our institution.

8.5.5 Clinical study


KAP values were recorded with the built-in KAP meter for 154 routine PCI
procedures performed in 1998 and 1999 with factory settings prior to optimiza-
tion. After optimization, 138 PCI procedures were recorded with the new
parameters in 2000 and 2001. Twelve operators, six radiologists and six
cardiologists, participated in the study.

- 36 -
Materials and Methods

In planned PCI procedures, normally for stable angina, a coronary


angiography had been performed at a separate session before the PCI
procedure and was thus not included in the dose measurements. However, a
large proportion of the procedures were performed because of unstable angina
or acute myocardial infarction. In these cases, a coronary angiography was
performed in the same session as the PCI procedure and was therefore
included in the dose measurements. Thus, the radiation dose can be expected
to be higher. These cases are labelled combined in the tables.
In order to evaluate clinical events during the clinical study, the local qual-
ity control database was searched for complications during the procedure or
until the discharge from the hospital. Small puncture site hematomas or the
evolvement of a myocardial infarction after PCI performed for acute myocardial
infarction were not considered as complications.

8.6 Statistical methods


In papers I and II, the VGA results were presented descriptively. Agreement
between observers was evaluated in paper I with (kappa) statistic analysis.
The angle measurements in paper I and IQF values in paper III were evalu-
ated using analysis of variance (ANOVA). In the instance when the ANOVA
yields non-significant results, no further testing is needed. When the ANOVA
was significant, post-hoc tests were conducted. In a post-hoc test, multiple
comparisons are made to find the pairs with significant differences. To correct
for the risk of false significance in the post-hoc test when doing several paired
comparisons, a correction is usually applied. In paper I, a Bonferroni correction
was applied in the post-hoc test. In paper III, Duncan's multiple range test was
used. These corrections are two of several methods available to correct for this
risk. In paper III, differences between observers were not evaluated statistically
and all variables were assumed to be independent.
In paper V, the KAP values and fluoroscopy time had a positively skewed
distribution and were thus analysed after logarithmic transformation (Altman
1991; Martin et al 1994). The means of these parameters can be expressed as
geometric means, which utilise this logarithmic transformation.
The material in paper V was also unbalanced with respect to combined
procedures in that the combined procedures showed higher total KAP than PCI
only. After optimization, a higher proportion of combined procedures was found
compared to before optimization because of changes in referral practice. To
correct for the unbalance a linear regression was applied with total KAP as the
dependent variable, before/after optimization as the independent variable and
combined procedures as a covariate.
A multivariate analysis was performed in paper V to evaluate the influence
of combined procedures, number of vessels treated, size of the patient and
number of stents placed.

- 37 -
Hkan Geijer

8.7 Ethical considerations


In paper I, ten patients were subjected to double examinations with a slight
increase in effective dose, about ten percent. Before including a patient into
that study, informed consent was obtained. The local research ethics commit-
tee also approved that part of the study. All other patient imaging was per-
formed as routine examinations. Also in the PCI study (paper V), all patient
imaging was performed as routine examinations.

- 38 -
Results

9 Results
9.1 Scoliosis radiography

9.1.1 Radiation dose


The results of measurements and calculations in the Alderson phantom are
shown in Table 2.
In paper II the exposure protocol was optimized, which reduced the dose
considerably. Table 3 shows the results for the different steps in the optimiza-
tion process. The effective dose was reduced to 0.047 mSv or 31% of the base-
line effective dose for digital exposure.
The ESD varied along the length of the phantom with low values along the
cervical and thoracic spine and high values along the lumbar spine and pelvis
(Figure 11). For a point approximation of ESD, the dose value at slice 25, ap-
proximately at the level of the navel, was chosen. This value was very close to
the mean ESD for the whole series.

Imaging technique Effective dose KAP ESD


Measured Monte Carlo
2
(mSv) (mSv) (Gy cm ) (mGy)
Screen-film 0.087 0.086 0.43 0.51
Digital exposure 0.16 0.15 0.87 0.90
Optimized protocol NA 0.047 NA 0.21
Digital fluoroscopy 0.017 0.015 0.097 0.11

Table 2. Measured and calculated phantom dose values. The ESD was meas-
ured with TLDs and included backscatter. The KAP for the optimized protocol
is not directly comparable and is thus not shown in this table.
KAP = kerma area-product. ESD = entrance surface dose. NA = not applicable.

- 39 -
Hkan Geijer

Optimization steps Tube Tube Dose Acquisition Scan Number KAP ESD Effective IQF
potential current request rate speed of images dose
-1
(kV) (mA) (s ) (cm/s) (Gy cm2) (mGy) (mSv)
Exposure, baseline 70 Falling load 50% 2 3.5 36 1.07a (100%) 0.90 (100%) 0.15 (100%) 52
1. Lowered dose request 70 Falling load 33% 2 3.5 36 0.67 (63%) 52
2. Increased pulse frequency 70 Falling load 33% 3 7.5 27 0.48 (45%)
and scan speed
3. Fixed mA 70 320 33% 2 3.5 36 0.45 (42%) 58
4. Higher kV 81 160 33% 3 7.5 27 0.25 (23%) 0.21 (23%) 0.047 (31%) 62
a
Fluoroscopy, baseline 56-72 Fluoroscopy Fluoroscopy 3 3.5 57 0.12 (100%) 0.11 (100%) 0.015 (100%) 82

- 40 -
Increased scan speed 56-72 Fluoroscopy Fluoroscopy 3 7.5 30 0.06 (50%) 0.043 (39%) 0.0067 (45%)

Table 3. Settings that were evaluated during the optimization process together with the resulting dose and image quality values. The
dose values and number of images refer to runs with the Alderson dose measurement phantom. IQF values were measured with the
contrast-detail phantom. The relative dose values refer to changes with respect to the baseline.
a = The KAP value is different between papers I and II, probably because of different collimation. Within this table, the change in KAP

is valid.
KAP = kerma area-product. ESD = entrance surface dose. IQF = image quality figure.
Results

2.5 Figure 11. Entrance surface dose


Exposure (baseline) with the Alderson phantom for
Optimized digital exposure (baseline), the
2.0 Fluoroscopy optimized protocol and for digital
Entrance surface dose (mGy)

fluoroscopy. The dosimeters were


placed at the dorsal aspect of the
phantom. Slice 11 is at the lower
1.5
cervical spine of the phantom and
slice 32 is at the lower part of the
pelvis. The point approximation of
1.0 ESD was chosen at slice 25, at the
level of the navel.

0.5

0.0
11 13 15 17 19 21 23 25 27 29 31
Phantom slice #

10
Figure 12. Contrast-detail curves
for the CDRAD phantom, mean
values for three observers. The
further down to the left the curve
lies, the better the image quality is.
= Digital fluoroscopy, IQF=88.
Diameter of target (mm)

= Digital exposure, IQF=64.


= Screen-film, IQF=47.
1

0.1
0.3 0.4 0.5 0.6 0.8 1 1.3 1.6 2 2.5 3.2 4 5 6.3 8
Depth of target (mm)

- 41 -
Hkan Geijer

KAP (Gy cm) Figure 13. Image quality figure (IQF)


0 0.2 0.4 0.6 0.8 1 1.2 versus kerma area-product (KAP) for
40
the evaluated settings during the
Lowered
optimization, mean IQF values for
dose Exposure three observers. The baseline value
request baseline for digital fluoroscopy is shown for
50
Increased Fixed comparison. A lower value of IQF
scan speed, mA correlates with better image quality.
high kV
60
IQF

70

Fluoroscopy
80 baseline

90

9.1.2 Image quality


Contrast-detail curves using the CDRAD phantom for screen-film, digital expo-
sure and digital fluoroscopy are shown in Figure 12. The relationship between
KAP and IQF as assessed with the contrast-detail phantom during the optimi-
zation is shown in Figure 13.
Because the VGA study was performed in several steps, these results are
described as paired comparisons. The corresponding IQF values are also given
for each comparison.
Digital exposure vs film: At level T3, digital exposure was superior in almost
all aspects in the VGA study. At levels T10 and L3, which were similar, con-
trast, noise and sharpness were graded somewhat lower but visualisation of
end plates, suitability for judging skeletal abnormalities and suitability for
angle measurements were graded somewhat higher than film. Overall image
quality was graded considerably higher than film. The IQF values were 64 for
digital exposure and 47 for film showing a lower image quality for the digital
exposure protocol.
Digital fluoroscopy vs digital exposure: In general, digital fluoroscopy was
graded lower compared to digital exposure for the same patient, with the least
difference in subjective suitability for angle measurements. Overall image
quality was also graded lower than digital exposure. The IQF values were 88 for
digital fluoroscopy and 64 for digital exposure showing an even lower image
quality for digital fluoroscopy.
Optimized protocol vs digital exposure: The optimized protocol was graded
slightly lower in all aspects compared to baseline digital exposure, with the
least difference in subjective suitability for angle measurements. The IQF val-
ues were 62 for the optimized protocol and 52 for digital exposure in this
study, showing a slightly lower image quality for the optimized protocol.

- 42 -
Results

9.1.3 Angle measurements


Using the angle phantom, ANOVA revealed a significant result for the two fac-
tors of method and observer (method: F = 5.8, p = 0.005; observer: F = 6.4, p =
0.003; method and observer: F = 2.3, p = 0.062). For the post-hoc test, a 95%
confidence interval of 0.38 was calculated. By using this confidence interval,
there were significant differences between observers 1 and 3 and between ob-
servers 2 and 3 for digital fluoroscopy. Likewise, there were significant differ-
ences between film and fluoroscopy and between exposure and fluoroscopy for
observer 3. Neither of the other comparisons was significant. Observer 3 and
digital fluoroscopy yielded smaller angle measurements than did the other two
protocols. Angles for the three observers are shown in Figure 14.
Using clinical images, there was no significant difference in angle
measurements when comparing the ten double exposed patients with digital
exposure and digital fluoroscopy (method: F = 1.1, p = 0.32; observer: F = 0.61,
p = 0.55; method and observer: F = 0.45, p = 0.64). Mean angles and
reproducibility of angle measurements are shown in Table 4.

22.0
21.8 Figure 14. Mean phantom angle meas-
urements obtained by three observers with
Mean angles (degrees)

21.6
each protocol.
21.4
= Observer 1.
21.2
= Observer 2.
21.0
= Observer 3.
20.8
20.6
20.4
20.2
20.0
Film Digital Digital
exposure fluoroscopy

Number of Mean SD for Intraobserver


Imaging technique patients Mean angle all observers difference
Mean SD
Screen-film 30 14.1 1.21 -0.02 1.58
Digital exposure 30 17.1 1.15 0.11 1.56

Digital exposure 10 14.8 0.87 -0.19 1.09


Digital fluoroscopy 10 14.4 0.89 0.02 1.03

Table 4. Means and SDs of Cobb angle measurements obtained with differ-
ent protocols. Three observers measured each angle twice each. The im-
ages for screen-film and digital exposure were obtained in different
patients and should not be assumed to have the same mean angles. The
ten double examined patients (digital exposure + digital fluoroscopy) are
shown in the two bottom rows.

- 43 -
Hkan Geijer

9.2 Flat-panel detector

9.2.1 Radiation dose and image quality


The tube charge and entrance dose values are shown in Table 5 together with
the resulting IQF values. This relationship is also shown graphically in Figure
15. The results for observer 1 and 3 (one radiologist and one medical physicist)
were constantly very close to the mean for all settings. Observer 2 and 4 (one
radiologist and one medical physicist) deviated considerably from the mean,
and because of this, the SD for all twelve readings was much larger than the
intra-observer SD for the three readings for each observer.

Imaging system Tube charge Entrance dose IQF SD


(mAs) (mGy)
Flat-panel detector
400-speed 10.3 0.44 30 4.0
600-speed 6.8 0.30 38 7.6
800-speed 4.5 0.20 43 9.1
Storage phosphor plates 13.8 0.64 40 6.8
Screen-film 40 1.7 35 8.1

Table 5. Tube charge, entrance dose and image quality for the
evaluated systems. The flat-panel detector was evaluated at
three different system speed settings. All exposure and dose
values are means for three images while the IQF values are
means of twelve readings for each setting (four observers and
three images).
IQF = image quality figure.

Entrance dose (mGy) Figure 15. Image quality


0.0 0.5 1.0 1.5 2.0 figure (IQF) versus entrance
15 dose for the four observers.
20
A lower value of IQF corre-
lates with better image
25 quality.
30 FP = flat-panel detector at
35 Mean different speed settings.
IQF

40 Observer 1
45 Observer 2
50 Observer 3
55 Observer 4
60
65 I I I I I
FP FP FP Storage phosphor Film
800 600 400 plates

- 44 -
Results

Imaging system Flat-panel detector Storage Screen-


400-speed 600-speed 800-speed phosphor plates film
Flat-panel detector
400-speed 0.025 0.00028 0.0027 0.15
600-speed 0.025 0.10 0.35 0.35
800-speed 0.00028 0.10 0.41 0.015
Storage phosphor plates 0.0027 0.35 0.41 0.080
Screen-film 0.15 0.35 0.015 0.080

Table 6. Individual p-values for the ANOVA comparing the flat-panel detector to
other detectors. Significant results are shown in bold.
FP = flat-panel detector at different speed settings.

The ANOVA analysis of IQF values yielded significant results (F = 5.3, p =


0.0011). The flat-panel detector at the highest dose setting (400-speed) had the
best image quality results. The post-hoc test showed significant results for the
comparisons with storage phosphor plates, with the flat-panel detector at the
intermediate setting (600-speed) and with the flat-panel detector at the lowest
dose setting (800-speed). There was also a significant difference between film
and the flat-panel detector at the lowest dose setting (800-speed) with film
being better. Neither of the other comparisons was significant. Individual p-
values are shown in Table 6.
In the parallax test, it was not possible to include the entire CDRAD phan-
tom in the field. Because of this, a complete IQF could not be calculated. In-
stead, we calculated a partial IQF, consisting of data from the eight rightmost
columns of the 15 in the phantom. The partial IQF was 17.7 (SD 1.9) when
centred in the middle of the phantom like the standard setting, 17.0 (SD 1.5)
when centred halfway between the middle and the lower right corner, and 17.2
(SD 1.9) when centred in the lower right corner. When the focus-detector dis-
tance was decreased from 110 to 100 cm, the partial IQF was 17.6 (SD 2.5).
No significant difference between the four groups was found with ANOVA (F =
0.34, p = 0.79).

- 45 -
Hkan Geijer

9.3 Coronary intervention

9.3.1 Radiation dose


Primary and scattered radiation at various settings using the Alderson
phantom are summarized in Figure 16. This graph shows the combined effect
of filtration, collimation, protective screens and angulation on the dose rates to
both patient and operator. As can be seen, large dose reductions can be
achieved. The operator dose rate (absorbed dose from scattered radiation) was
especially sensitive to the changes, with a maximum reduction to 4% of the
original value. With angulation, the scattered dose rate to the operator instead
increased to 240% of the original value.

2.5

2.0
Relative dose rate

1.5
Calculated
effective
1.0 dose rate
KAP rate

0.5
Operator
dose rate

0.0
+angulation Start +filtration +collimation +screens
LAO caud
Combined actions

Figure 16. This graph shows the combined effect of filtration, collimation,
protective screens and angulation on the dose rates to both patient and opera-
tor.
The second column, "Start", shows the normalised dose rates for effec-
tive dose, KAP rate and operator dose rate for a neutral starting position
(PA projection, 18 cm image intensifier size and no additional filtration).
The next column to the right shows the effects of additional filtration (the
thinnest extra filter with 1.5 mm Al and 0.1 mm Cu).
In the fourth column, the X-ray field has been collimated 2 cm at each
side.
In the fifth column, screens have been inserted between the patient and
the operator.
If a rather heavy angulation and rotation is applied (LAO 45
degrees/caudal 35 degrees) as is shown in the first column, KAP and
operator dose rate instead rise considerably.

- 46 -
Results

Effective dose in a simulated PCI procedure: The effective dose for a stan-
dard PCI procedure before optimization was 13 mSv. After optimization, the
measured effective dose was 4.6 mSv, showing a reduction to 35% of the origi-
nal effective dose. Using Monte Carlo techniques, the calculated effective dose
was 8.1 mSv before and 3.5 mSv after optimization, a reduction to 43% of the
original value. KAP values, ESD and operator dose from these procedures are
shown in Table 7.

Series Rotation Angulation Type Median KAP ESD Operator dose


Number KAP in
patient
population
2 2
(degrees) (Gy cm ) (Gy cm ) (mGy) (Gy)
before after before after before after
#1 RAO 30 cran 20 fluoroscopy 12.1 2.8 13 1.8
acquisition 5.9 3.7 4.6 1.9
total 18.0 6.5 296 116 17 3.6

#2 RAO 30 caud 20 fluoroscopy 11.8 2.4 12 0.3


acquisition 7.0 4.3 5.7 0.4
total 18.8 6.7 311 125 18 0.7

#3 LAO 30 caud 20 fluoroscopy 11.9 3.8 14 0.5


acquisition 4.5 3.1 4.4 0.2
total 16.4 6.9 261 137 19 0.7

#4 LAO 30 cran 20 fluoroscopy 11.8 2.7 21 0.6


acquisition 7.1 5.0 10 0.3
total 18.9 7.8 323 147 32 0.9

Total fluoroscopy 48 47.5 11.7 61 3.2


acquisition 24 24.5 16.1 25 2.8
total 72 72.1 27.8 1,191 525 85 6.0

Table 7. Radiation dose with an Alderson phantom for a standardized PCI procedure before
and after optimization. The operator dose was measured in chest level at the normal location of
the operator in the room. The apparent inaccuracies in the sums are due to rounding effects.
KAP = kerma area-product. ESD = entrance surface dose.

- 47 -
Hkan Geijer

40

NC
45

LP
LP HP
12.5 NP
50
LP-R
LC
IQF

55

60
LC-R

65
LP12.5-R

70
0 20 40 60 80 100
KAP rate (mGy cm2/s)

Figure 17. Image quality (IQF) versus KAP rate for equipment settings in a cardiac catheterisa-
tion laboratory. The arrows show changes from the original low-dose settings LC and LP that
were evaluated during the optimization process. All settings starting with L use the thickest
extra filtration (1.5 mm Al + 0.4 mm Cu). All settings starting with N use the thinnest extra
filtration (1.5 mm Al + 0.1 mm Cu). The setting starting with H is without extra filtration. C
denotes continuous and P pulsed fluoroscopy.
Both reducing the pulse frequency from 25 to 12.5 pulses per second (LP LP12.5), and
reducing the image intensifier dose request to half (LP LP -R), reduced the KAP rate to
about half but decreased image quality slightly.
When both these measures were combined (LP12.5-R), image quality was heavily af-
fected.
When the dose request in continuous fluoroscopy was reduced by 65% (LC LC -R) the
KAP rate was reduced accordingly but image quality also suffered.
The net change implemented during the optimization was from setting NP to LP-R.

9.3.2 Image quality


The resulting IQF values with the contrast-detail phantom can be divided into
two parts: changes in equipment settings and changes due to patient and
operator factors. These are shown in Figure 17 and Figure 18, respectively.

- 48 -
Results

40

NCcoll
NC
45

LP
NClight
50
IQF

55

LPthick
60

65

70
0 20 40 60 80 100
2
KAP rate (mGy cm /s)

Figure 18. Image quality (IQF) versus KAP rate for changes due to patient and operator
factors in a cardiac catheterisation laboratory. The arrows show changes from the original
settings that were evaluated during the optimization process. LP = pulsed fluoroscopy with
extra filtration 1.5 mm Al + 0.4 mm Cu. NC = continuous fluoroscopy with extra filtration 1.5
mm Al + 0.1 mm Cu.
Collimation: 2 cm collimation at each side reduced the KAP rate to less than half
without any substantial change in image quality (NC NCcoll).
Ambient lighting: Increased lighting in the room (NC NClight) reduced the image
quality slightly.
Object thickness: When the acrylic thickness was increased from 15 to 20 cm (LP
LPthick), image quality was considerably reduced while the KAP rate increased.

9.3.3 Patient dose data


Before optimization, the mean total KAP was 93.6 Gy cm2. After optimization,
the corresponding value for total KAP was 69.1 Gy cm2. Detailed dose data are
shown in Table 8. Using linear regression to correct for the different
proportions of combined procedures before and after optimization, the
fluoroscopy KAP was reduced significantly to 45% of the original value and the
total KAP was reduced significantly to 67% of the value before optimization.
The digital acquisition KAP and fluoroscopy time did not change significantly.

- 49 -
Hkan Geijer

Period Fluoroscopy Digital Total KAP Calculated Fluoroscopy Procedure Number of Number of
Statistic KAP acquisition effective time Time procedures stents per
KAP dose with including procedure
conversion stent
factors placement
2
(Gy cm ) (Gy cm 2) (Gy cm 2) (mSv) (min) (min)
PCI only
Before (n=94)
Mean 55.9 23.3 79.2 13.5 50.0 84 (89%) 1.28
Median 39.9 19.2 61.8 11 9.8 42.5
Third quartile 92.9 17.1
After (n=59)
Mean 33.8 22.4 56.2 16.4 48.2 53 (90%) 1.54
Median 16.5 20.1 41.3 6.8 12.1 44.0
Third quartile 66.5 20.5
Combined proce-
dures
Before (n=60)
Mean 71.7 44.4 116 15.3 62.6 52 (87%) 1.12
Median 57.5 38.7 97.1 18 12.8 57.0
Third quartile 146 19.4
After (n=79)
Mean 36.2 42.5 78.7 14.9 53.0 74 (94%) 1.37
Median 24.2 35.0 61.4 10 12.9 48.0
Third quartile 99.8 19.4
All patients
Before (n=154)
Mean 62.0 31.5 93.6 14.2 55.0 136 (88%) 1.21
Median 46.0 24.8 72.6 13 11.2 48.5
Geometric mean 47.5 25.4 75.7 11.4
Third quartile 121 18.4
After (n=138)
Mean 35.2 33.9 69.1 15.6 51.0 127 (92%) 1.44
Median 22.3 28.2 55.4 9.2 12.6 47.0
Geometric mean 22.0 28.3 54.6 12.0
Third quartile 82.6 19.6
Ratio after to
before optimiza- 0.45 0.99 0.67 1.03
a
tion
95% confidence
0.36-0.55 0.86-1.13 0.57-0.78 0.87-1.22
interval

Table 8. Dose and clinical data before and after optimization of settings for coronary inter-
vention.
KAP = kerma area-product.
a The ratio after to before optimization was corrected for the imbalance in number of com-

bined procedures. A ratio of 1 indicates no change during the optimization. A ratio of 0.67
indicates a reduction to 67% of the original value.

- 50 -
Results

The multivariate analysis showed that the total KAP increased with a mean
factor 1.42 (95% confidence interval 1.22-1.64) for a combined procedure
compared to PCI only. The total KAP increased with a factor 1.14 (95%
confidence interval 1.09-1.19) for each cm increase of the equivalent diameter.
There were three complications (1.9%) noted before optimization: one
balloon rupture during stent implantation, one air embolisation in the right
coronary artery and one dissection in an obtuse marginal that was stented. All
these patients recovered uneventfully. After optimisation, there were six
complications among the 138 patients (4.3%); two puncture site
pseudoaneurysms that were treated with ultrasound-guided compression, one
large dissection that was stented with five stents, still open without restenosis
at angiography two months later, and one abrupt closure which ended with a
myocardial infarct. There was also one case of diminished flow after balloon
dilatation where the patient recovered uneventfully, and finally an old
occlusion that couldnt be opened during infarct-PCI.
Altogether there were twelve operators included in the study; six cardiolo-
gists and six radiologists. Of these, three cardiologists and three radiologists
performed cases both before and after the optimization. These six performed
127 out of 154 procedures (82%) before and 95 out of 138 procedures (69%)
after optimization. Cardiologists and radiologists performed about equal
amounts of procedures. KAP values for these six operators are shown in Figure
19 and in greater detail in Figure 20.

- 51 -
Hkan Geijer

Figure 19. Mean total KAP before


160
and after optimization for the six
operators who participated in both
140 parts of the study. The number of
procedures performed during each
120 period is shown for each operator.
Mean KAP (Gy cm)

A (26/14)
100 B (23/5)
C (20/32)
80
D (16/20)
60 E (40/19)
F (2/5)
40

20

0
Before After

Figure 20. Box-and-whisker plots of the total KAP before and after optimization for the six
operators who participated in both parts of the study. The number of procedures for each period
is shown over the bars.

- 52 -
Discussion

10 Discussion
The results of this project indicate that it is possible to reduce radiation dose
further, even in a modern digital radiology department. In scoliosis radiography
it was possible to reduce the effective dose to one tenth using pulsed fluoro-
scopy instead of digital exposure. By optimizing the digital exposure protocol
the effective dose was reduced to about one third. By exchanging storage
phosphor plates or film for a flat-panel detector it was possible to both reduce
dose and improve image quality. In coronary intervention, finally, the
fluoroscopy dose rate could be reduced so much that the total KAP was
reduced to about two-thirds.
This Discussion begins with specific parts for the three main studies on
scoliosis radiography, flat-panel detector and coronary intervention. Then, the
statistical methods are covered followed by discussions on radiation dose and
image quality. Finally, the subject of optimization methods is discussed.

10.1 Scoliosis radiography


The digital exposure protocol is feasible, albeit with a high effective dose, twice
that of the screen-film method. Images are subjectively of good quality with a
consistent exposure along the entire length of the spine without under- or
overexposed areas. By using either digital fluoroscopy or the optimized proto-
col, the radiation dose is reduced considerably, although with a concomitant
reduction in image quality. Bearing in mind that the major reason for scoliosis
radiography is angle measurements, this trade-off in image quality should be
acceptable while being in accordance with the principles of optimization and
ALARA.
Storing of fluorographic images such as in the digital fluoroscopy protocol
is similar to the use of a 100 mm camera when using film-based systems. Us-
ing a 100 mm camera, a dose reduction of 80-90% compared to screen-film has
been reported in barium meal examinations (Martin and Hunter 1994).
A problem with the present reconstruction technique is that a rather large
overlap between images is needed, which increases radiation dose. An alterna-
tive approach is to attach two or three storage phosphor plates to a stand,
slightly overlapping, and then image the whole spine in one exposure, just like
the traditional screen-film method. The resulting images can then be stitched
together in a workstation (Dewaele et al 1999).
The radiation dose in scoliosis radiography has previously been lowered in
various ways (Nash Jr et al 1979; Ardran et al 1980; Gray et al 1983). Changes
in the screen-film system at conventional imaging can reduce the dose consid-
erably (Hellstrm et al 1983). PA instead of anteroposterior (AP) projection re-
duces the dose to the breast (Levy et al 1996) and other sensitive organs at the
cost of a somewhat higher bone marrow dose (Fearon et al 1988). PA projection
has been reported to be superior in image quality to AP projection (Frank et al
1983), which is now considered obsolete in scoliosis radiography (Almn et al
1996). The normal lumbar lordosis should be better imaged in PA projection
(Tsuno and Shu 1990) while the thoracic spine, especially if kyphotic, might be

- 53 -
Hkan Geijer

more difficult to image in a PA projection. The geometry in the present scan-


ning method is somewhere between the traditional AP and PA projections since
the X-ray beam is nearly parallel in the cranio-caudal direction. An under-
couch tube position, when used with a tabletop, is valuable because the table-
top attenuates considerably. We have measured attenuation as high as 30%
with the parameters used in this study. Such an attenuator between the pa-
tient and the image receptor should be avoided because this increases the dose
to the patient. Increasing X-ray beam filtration is another way to lower the dose
(Gray et al 1983). With the transition to digital imaging a further reduction in
radiation dose has been forecasted (Kogutt et al 1989; Kalmar et al 1994;
Stringer et al 1994).
The digital exposure images were subjectively rated as better than the
screen-film scoliosis images, despite visible noise, especially when magnified.
This was also reflected in the visual grading part of the study, in which the
digital exposure images were rated as superior to the screen-film images in
several aspects. The optimized protocol was graded inferior to the baseline
digital exposure protocol but superior to the digital fluoroscopy protocol, which
is to be expected. Somewhat surprising is the fact that the fluoroscopy protocol
was considered clinically feasible despite the very low radiation dose.
Using the angle phantom, there was a slightly larger variability for digital
fluoroscopy compared to screen-film and digital exposure. However, the vari-
ability in absolute terms was very low in the angle phantom measurements
with a mean angle SD of only about 0.5, compared with about 1.0 on the pa-
tient images. This means that a small, probably clinically insignificant differ-
ence might turn out statistically significant. In clinical practice, differences of
up to 10 between measurements have been required to indicate a real change
in the scoliosis angle (Carman et al 1990). No significant difference between
measurements was found on the patient images.
This study points to the possibility of combining two methods (Kalmar et al
1994). The high-dose protocol with best image quality could be used for the
initial examination, in which it is important to evaluate underlying abnormali-
ties, whereas a low-dose protocol with inferior image quality could be used for
the follow-up examinations, in which the focus is angle measurements. Then,
the patient is not subjected to more radiation exposure than absolutely neces-
sary.

10.2 Flat-panel detector


The flat-panel detector was shown to have a superior image quality compared
to storage phosphor plates and a screen-film system. It is even possible to both
reduce dose and improve image quality at the same time. A comparable image
quality was reached at less than half the entrance dose compared to storage
phosphor plates and at about one fifth of the dose compared to the screen-film
system. Several groups have shown similar results; improved image quality at
unchanged dose, or reduced dose with the same image quality, using the
Trixell detector (Chaussat et al 1998) in Philips version (Bury et al 1998, 2000;
Neitzel et al 2000) as well as Siemens (Vlk et al 1997, 2000; Strotzer et al
1998a-b, 2000; Spahn et al 2000). Also the Canon (Garmer et al 2000), General
Electric (Aufrichtig 1999; Aufrichtig and Xue 2000; Granfors and Aufrichtig

- 54 -
Discussion

2000; Floyd Jr et al 2001; Chotas and Ravin 2001) and Sterling (Piraino et al
1998) flat-panel detectors have been evaluated with similar results.
This beneficial balance between dose and image quality is largely due to
the higher DQE which has been reported at 0.6 to 0.75 (Chaussat et al 1998;
Bury et al 2000; Antonuk et al 2000; Granfors and Aufrichtig 2000; Spahn et al
2000; Floyd Jr et al 2001). This is about twice that of storage phosphor plates
at DQE = 0.3 (Kengyelics et al 1998) and at least twice that of screen-film
systems (Neitzel et al 2000). For the General Electric type of flat-panel detector,
the exposure needed for equal image quality with a contrast-detail phantom
has been predicted to be only 30% of the screen-film exposure (Aufrichtig
1999).
The Cesium Iodide (CsI) scintillator is a very important part of the flat-
panel detector (Chaussat et al 1998) and it is definitely contributing to the high
DQE. It has a geometric configuration of small channels acting as light guides,
thus preventing geometric light spread even at a large thickness. The X-ray
absorption is superior to selenium at energies higher than about 35 keV, and
the emitted light has a wavelength that matches the spectral sensitivity of the
amorphous silicon photodiode. Other reasons for the high DQE are the small
pixel size and the low electronic noise of the signal readout chain (Spahn et al
2000).
The high DQE of the flat-panel detector is valid only for higher
(radiographic) doses such as those used in this study (Siewerdsen et al 1998;
Antonuk et al 2000). For lower doses such as in fluoroscopy, the DQE falls
rapidly due to electronic noise in the detector (Chotas et al 1999; Maolinbay et
al 2000). The role of the flat-panel detector in fluoroscopic imaging is evolving
rapidly, but it has not been evaluated in this project.
This study points to an important way of reducing radiation dose by
exchanging the detector system. The recently available flat-panel detector
should enable important dose reductions in high-dose procedures such as
lumbar spine radiography and intravenous urography, and it is one of the most
important new devices that has been introduced in diagnostic radiography
recent years. However, it should be noted that it is not enough just to install a
new detector. Even using an efficient flat-panel detector, an optimization to the
lowest dose with acceptable image quality should still be done.

10.3 Coronary intervention


This study resulted in a large and significant reduction of the KAP for clinical
PCI procedures. The total KAP was reduced to 67% of the original value by
changing the settings for fluoroscopy in a cardiac catheterisation laboratory.
All dose reduction can be attributed to decreased fluoroscopy KAP since the
digital acquisition KAP and fluoroscopy time did not change during the
optimization process. The fluoroscopic image after optimization was initially
considered noisy, but was quickly accepted. Furthermore, the operator can
whenever he or she wishes increase the dose in two steps simply by pushing a
button at the tableside console.
When compared to other studies (Va et al 1995a, 2001; Broadhead et al
1997; Bakalyar et al 1997; Zorzetto et al 1997; Betsou et al 1998; Padovani et
al 1998; Bernardi et al 2000; Fransson and Persliden 2000; Katritsis et al

- 55 -
Hkan Geijer

2000; Widmark et al 2001), the mean KAP values in the clinical part of this
study are comparable; before optimization they are in the higher range, and
after optimization they are in the lower range.
There was a very large range in total KAP with a fiftyfold difference between
the lowest and the highest total KAP. This reflects the widely varying clinical
conditions, from a quick single-vessel PCI with direct stenting and only one
balloon inflation to a complex multi-vessel case, perhaps with dissection and
other complications.
There was also a large difference in KAP between operators. This difference
between operators has been noticed before (Clark et al 2000). Especially in
complicated cases, there can be a large difference in dose depending on
whether the operator is a high-dose or low-dose operator.
There has been a recent discussion on diagnostic reference levels for
different radiological procedures in the EU (Commission of the European
Communities 1996; The Council of the European Union 1997). If a sample
from an institution exceeds the reference dose level, an investigation should be
started and correction measures taken. Provisional quality criteria for coronary
angiography and reference levels for diagnostic and interventional procedures
have been proposed (Padovani et al 1998; Widmark et al 2001). The proposed
reference KAP levels for PCI are 120 Gy cm2 with a fluoroscopy time of 24 min
by Padovani et al and 90 Gy cm2 by Widmark et al. Our results for PCI are well
below these figures, even before optimization. However, the concept of reference
levels is difficult to adopt to interventional radiology because of the extreme
variability of procedures (Marshall et al 2000). Still, coronary angiography and
PCI are fairly standardized for being invasive procedures.
Analogously, one might conceive the concept of reference levels for
individual operators. If an individual operator repeatedly exceeds the reference
level, an investigation should start. Often, the solution would probably be
education and intensified training to make the operator aware of the problem
and show him or her ways to reduce the dose.
The dose issue in interventional radiology has gained increased attention
recent years with reports such as ICRP 85 (ICRP 2000) and papers describing
the risks of skin injuries (Koenig et al 2001a-b). The ESD reported in this study
for a standardized PCI procedure, 1.2 Gy before and 0.53 Gy after, is well below
the risk of skin erythema, which starts at about 2 Gy. The reports above only
deal with the deterministic risks, but the stochastic risk for induction of
malignancy must not be forgotten, although many PCI patients are so old and
at high risk from their underlying disease that they run a small risk of
developing a radiation-induced malignancy. However, another extremely
important reason to keep doses low is the occupational aspect. Operators and
staff receive significant doses while performing these procedures, so there is a
double gain in reducing dose. This ought to make operators more willing to
adopt dose-saving techniques, but that is not always the reality. All changes
that reduce the patient dose are also beneficial for the operator (Marshall and
Faulkner 1992). Most dose reductions impair image quality more or less, but
some actually increase image quality.

- 56 -
Discussion

Modern angiographic equipment is highly advanced but increased attention to


dose reduction is very important:
It is vital for the operator to be able to change fluoroscopy dose-rate at the
table, otherwise it impossible to use such an ultra low-dose fluoroscopy
setting as in the present study. Without this possibility, the present dose
reduction had not been possible.
Of course the lateral projection should be performed with the image
intensifier at the same side as the operator. Still, this is not always the
case.
It should also be possible to adjust collimation and place the wedge filters
without using fluoroscopy, instead using a last image hold (LIH) image.
The old and well-known concept of compression is not feasible in cardiac
catheterisation, but in abdominal angiography it can give considerable
dose reductions and also improve image quality. Unfortunately the equip-
ment available for abdominal compression in angiography is not very well
developed.
The filtration of the beam consisted in this study of aluminium and cop-
per. New filter materials such as yttrium, gadolinium and ytterbium (Kohn
et al 1988; McParland and Boyd 2001) have been tested, but aluminium
and copper are still the most common filter materials.

Education and training of operators has also gained increased attention


(Johnson et al 2001). A previous study (Pitney et al 1994) has shown good
effect of training and education. The present study included some training of
operators and staff but did not evaluate this effect separately. Evaluation was
focused mostly on changes in equipment set-up and not in user behaviour. It
seems that the use of collimation did not change much since the digital
acquisition KAP was unchanged in the clinical study. Furthermore, the
reduction of fluoroscopy and acquisition KAP in the clinical study was less
than in the phantom simulation where a collimation of about 7 mm was
introduced at each side.
One possible future development is the so-called region of interest (ROI)
filtration (Labbe et al 1994; Rudin et al 1996; Kezerashvili et al 1997). The ROI
filter is a semi-transparent X-ray beam filter that is thicker at the edges of the
image and thinner in the centre. The less filtered beam in the centre of the
image is less noisy which gives a better image quality. In the periphery, where
image quality requirements are less strict, the signal is lower and the noise is
higher. In the computer system the grey scale in the different parts of the
image is adjusted to give a normal appearance. There is thus a varying image
quality in different parts of the image; high in the centre where the object of
interest (eg. stent or catheter tip) is normally placed, and lower in the periphery
which still has to be included in the image for orientation purposes. Altogether
this could lead to dose savings.
Another issue is the monitoring of dose. On-line monitoring with
continuous display in the catheterisation laboratory is becoming a legal
requisite (The Council of the European Union 1997; Johnson et al 2001) and it
will therefore become more prevalent in the future. Preferably both KAP and
ESD should be displayed since ESD is the major determinant for skin damage
and the conversion of KAP data to ESD in a population is unreliable (van de

- 57 -
Hkan Geijer

Putte et al 2000). Once displayed, it should then be easy to transmit the dose
data automatically to the radiological information system (RIS) where it can be
used for follow-up. Unfortunately ESD is very difficult to measure accurately in
a clinical situation with varying projections (Va et al 2001).
When performing simulated PCI procedures on an Alderson phantom,
there was a reduction of the fluoroscopy KAP value to 25% and the digital
acquisition KAP value to 66%, resulting in a reduction of the total KAP value to
39% of the value before optimization. The change in digital acquisition KAP can
probably be attributed to increased collimation (about 7 mm on each side, see
Figure 10 on page 35) since no other parameters were changed in the
acquisition settings. This together with the fact that the acquisition KAP wasnt
reduced in the clinical study shows that the clinical dose can probably be
reduced further through increased use of collimation in PCI.

10.4 Statistical methods


The statistical methods that have been used in this project are mostly uncon-
troversial and are thus not mentioned further. Some methods, however, merit
further discussion.
Logarithmic transformation of dose data: Data from a clinical dose survey
tend to have a positively skewed distribution (Martin et al 1994). To correct for
this, a logarithmic transformation can be applied to make data approach a
normal distribution, which is a requisite for using many statistical methods.
The calculations can be performed on the log data and the results are then
transformed back to the original scale. This approach was used in paper V.
Kappa analysis: In paper I, kappa () statistic analysis was used in the Vis-
ual grading analysis. This statistical method is a measure of agreement be-
tween observers. The maximum value is 1.00 when agreement is perfect,
whereas a value of zero indicates no agreement better than chance. statistic
analysis is of limited value when most of the gradings for all observers accu-
mulate at only one level (eg, are equal). Then, calculations yield low values
despite excellent agreement. In the extreme case when all the grades on all the
images for all observers are identical, the value cannot be calculated owing to
a division by zero. Kappa analysis was thus not very beneficial in this setting
where the main interest was the difference between methods.
Ordinal data: In VGA, it is quite common to set a numeric value to each
level, for example much worse = 1, worse = 2, equal = 3, etc in the visual
grading scale and then use the figures to calculate mean values and perform
statistical analysis (Manninen et al 1984; Mnsson 1994). This even seems to
be almost accepted practice. However, this methodology violates one basic
principle in statistics because the levels are ordinal data with no assumption of
equal steps between the different levels. Consequently, it is questionable to use
these data as continuous data in the analysis (Altman 1991). In papers I and II
a descriptive presentation of the material is given instead.

- 58 -
Discussion

10.5 Radiation dose


The dose reduction activities described in this document have resulted in large
and important reductions of the radiation dose, both to patients and staff. If
these principles are applied more widely, a substantial reduction of the collec-
tive dose from diagnostic radiography could result. As radiation dose is linked
to a cancer risk, a decreased dose will also reduce the risk of cancer.
In the present project, many dose values were recorded as KAP values.
These are easy to obtain and they are valuable indicators of the radiation expo-
sure to the patient. Conversion factors from KAP to effective dose for different
energies and areas of the body can be found in the literature (Hart et al 1994).
The effective dose can also be calculated with Monte Carlo methodology using
KAP as input value. KAP lends itself especially for dose surveys in a clinical
population where the easy handling of the equipment is valuable. When per-
forming an optimization of a system, perhaps exposing multiple images of the
same phantom with different settings, KAP can be difficult to use because of
the sensitivity to collimation. A slight variation in collimation between images
will influence dose falsely. Instead, other dose values such as ESD might be a
better choice.
KAP gives a good value of the stochastic risk of induction of malignancy.
However, it is of less value in showing the risk of deterministic effects such as
skin burns (ICRP 2000). Then, ESD gives a much more reliable estimation of
this risk.
KAP values are more sensitive to increases in kV level and filtration than
effective dose is, and might thus exaggerate the dose reduction of increased kV
or filtration. The increased penetration of the X-ray beam due to the spectral
change deposits more of the radiation dose deeper in the patient which is not
accounted for when measuring KAP or entrance dose (Shrimpton et al 1988).
However, this effect is accounted for in the effective dose.
Measuring organ doses in an Alderson phantom for calculation of effective
dose is considered to give accurate results. It is very time-consuming because
of the large number of TLDs that have to be prepared, loaded, unloaded and
read. In the studies for papers I and IV, each phantom was loaded with about
100 TLDs in various organs. The Monte Carlo simulation for paper I yielded
almost exactly the same effective dose as the phantom measurements, and
Monte Carlo methodology seems to be a very fruitful way of assessing effective
dose. It is certainly much easier to perform than measuring absorbed dose in a
phantom.
When calculating effective dose for the PCI procedures using the Alderson
phantom, there was a disagreement between measured and simulated effective
doses. This discrepancy has not been fully explained. However, both methods
include several estimations and uncertainties that are especially prominent
with the very narrow beam and oblique projections in PCI. In the situation with
an Alderson phantom packed with dosimeters it was difficult to estimate how
much of an organ was included in the primary X-ray field. An organ is usually
represented by only a few TLDs, and the dose is very dependent on whether a
TLD is included in the primary beam or not. Similarly, when doing Monte Carlo
simulations it can be difficult to ascertain the correct geometry, eg. focus-to
skin distance and field size, which are used as input values.

- 59 -
Hkan Geijer

The thickness of the patient influences KAP and ESD profoundly, and it is
thus vital to record patient height and weight in a patient survey. When re-
ducing the sample size to eg. ten to 20 patients as has been advocated by the
EU (Commission of the European Communities 1996), even one or two individ-
ual patients may influence the mean value considerably. To minimise this ef-
fect, it is often recommended that only patients within certain weight limits
such as 60 to 80 kg are included in the study. A drawback of this concept is
that about half of the subjects will be excluded, which decreases the statistical
power of the study. An alternative approach is to use the information about
patient size (Lindskoug 1992). It has been shown that there is a linear
relationship between the logarithm of entrance dose and equivalent diameter
(Martin et al 1994; Mooney and Thomas 1998). It should thus be possible to
measure doses on all patients and then apply a correction factor for patient
size (Chapple et al 1995).

10.6 Image quality


Contrast-detail phantoms: In the present project, a contrast-detail phantom was
used very frequently. Such a phantom allows an evaluation of the whole imag-
ing chain including the observer and can be seen as a substitute for clinical
images. Advantages are that the contrast-detail phantom is easy to use, no
humans are subjected to potentially harmful radiation and that all settings can
be well controlled. With the CDRAD phantom a numerical value of image
quality (IQF) is also produced and with this value, differences in image quality
between imaging systems and settings can be appreciated. With experienced
observers, even small differences in image quality can be shown. Previous
studies have shown good correlation between phantom studies and studies
based on assessment of clinical images (Va et al 1995b; Martin et al 1999b).
An advantage of the CDRAD phantom is that the targets are randomly
distributed in the squares making the test a four-alternative forced choice.
During the image quality evaluations it was noticed that the targets were
located correctly more often than the observer believed he could see the object.
Any errors due to guessing are accounted for when the readings are corrected
according to the user manual. One drawback is the variability between
observers (ICRU 1996) as was shown in paper III with a large difference in IQF
between observer 2 and 4. It seems that observers use different levels of
confidence in deciding whether a target can be seen or not. However, the trend
between the different settings was very similar for all observers as is shown in
Figure 21. Compared to other physical measurements of image quality, the
contrast-detail phantoms come closer to the real-life situation when the
radiologist is appreciating image quality in a clinical radiograph.
The large inter-observer variation raises the question whether some kind of
calibration of the observers might be appropriate. In Figure 22 the same data is
shown as in the previous figure, but now each observer has been calibrated to
his own mean result for the whole study, showing a much lower variability.
This of course would increase the statistical power of the readings.

- 60 -
Discussion

Entrance dose (mGy) Figure 21. Individual Im-


age quality figures (IQF)
0.0 0.5 1.0 1.5 2.0
at different settings for
15
the flat-panel detector
20 compared to storage
25 phosphor plates and
30 screen-film, mean of three
images for each setting.
35
Mean (SD) There are large differ-
IQF

40 Observer 1 ences between observers


45 Observer 2 showing up in the large
Observer 3 SD of the mean IQF, but
50 Observer 4
the level for each observer
55
is consistent compared to
60 the other observers.
65 I I I I I FP = flat-panel detector at
FP FP FP Storage phosphor Film different speed settings.
800 600 400 plates

Entrance dose (mGy) Figure 22. Individual Im-


age quality figures (IQF),
0.0 0.5 1.0 1.5 2.0
shown as delta () values
-25 from each observers
-20 mean for the whole study.
-15 The overall variation
-10 shown as SD of the mean
IQF is considerably re-
-5
duced.
IQF

Mean (SD)
0 Observer 1 FP = flat-panel detector at
5 Observer 2 different speed settings.
Observer 3
10 Observer 4
15
20
25 I I I I I
FP FP FP Storage phosphor Film
800 600 400 plates

Parallax effect: The X-ray beam is perpendicular at the centre of the


CDRAD phantom but divergent at the edges. This might impede the recognition
of the peripherally located holes, eg. in the high frequency part of the phantom
that is at the lower right corner (Neitzel et al 2000). However, the study in
paper III did not show any significant change in scores when the phantom was
moved to let the central beam enter at the lower right portion of the phantom.
This would mean that it is possible to get accurate readings with the phantom
placed in the central part of the X-ray field. This also increases repeatability
between images since it is much easier to place the phantom in the middle of
the X-ray field rather than aiming for an unspecified part closer to one corner.
For an evaluation of a clinical installation it is important to use the contrast-
detail phantom with geometry and other settings that are as close as possible
to a normal examination.

- 61 -
Hkan Geijer

VGA: Visual grading analysis is valuable in comparing image quality for


different systems or settings that are believed to be more or less equal in image
quality. Then, it is very sensitive in showing which system or setting is the
best. However, VGA is not very good to use when optimizing a system to a
lower dose with a consequently reduced image quality. The only answer from a
VGA study comparing an optimized setting to the baseline will be that the op-
timized images are inferior in quality compared to the baseline, and that will be
no surprise! Numerical values from a VGA study are questionable, see chapter
10.4 for a discussion on this topic. When comparing an optimized setting to the
baseline, it is probably more rewarding to use an image criteria score such as
the EU quality criteria (Commission of the European Communities 1996), but
this topic has not been evaluated in this project.
Simply comparing images and saying which is best might be seen as an
extreme variant of VGA. It seems as if radiologists are quite good at doing this.
In a recent study, two radiologists were able to sort randomly presented images
with varying mAs according to their image quality (amount of noise) despite
very small differences between the individual images (Persliden et al 2001).
ROC: Several studies have been conducted with Receiver operating char-
acteristics (ROC) analysis (Vlk et al 1997; Strotzer et al 1998b, 2000) and this
method was recommended in a report by the International Commission on
Radiation Units and Measurements (ICRU 1996). ROC is a powerful method for
evaluating image quality but also very time-consuming. It has not been used in
this project. Tingberg argues that the easier methods VGA or ICS give similar
results and thus can be used instead (Tingberg 2000). ROC has similarities to
a 2-alternative forced choice test (Burgess 1995), which is a variant of that
employed in the CDRAD readings.
In several studies comparing various types of digital imaging and film, the
digital images have been printed on film using a laser printer. By doing this,
the full potential of the digital images is not used since the ability to view and
adjust image settings directly at a monitor is one of the major advantages of
digital imaging (Ludwig et al 2000). Digital images should be viewed at a
monitor. Still, when optimizing a system, image quality should be evaluated at
the place where image viewing normally takes place. If an institution uses
printed laser films, then there might be no other choice than to use these films,
keeping in mind that the potential of digital radiography is not fully exploited.
Physical measures of image quality such as MTF or DQE have not been
used in this project. MTF is a way to describe the spatial resolution of an im-
aging system; it describes the transfer of high-contrast objects. Digital process-
sing can elevate MTF considerably and it is thus not a reliable descriptor of the
imaging system. The use of the MTF in characterising digital X-ray systems has
consequently been questioned. Moy argues that the DQE is a much better
parameter for describing image quality in digital radiography (Moy 2000). Re-
cent years there has been a growing acceptance that it is the low-contrast
resolution that is the most important descriptor of a digital system (Laskey et al
2000), since digital radiography is noise-limited (Cowen 1991). Thus, DQE
should be a better parameter. Unfortunately it is difficult to measure.

- 62 -
Discussion

10.7 Optimization methods

10.7.1 General considerations


Several clinical surveys have shown that image quality and radiation dose are
not very well correlated when comparing different institutions (Geleijns et al
1993; Leitz et al 1993). This poor relationship points to the possibility of
optimization. If you have both a high dose and a poor image quality the proce-
dure is certainly un-optimized. This possibility is targeted by the EU through
the concept of diagnostic reference levels (The Council of the European Union
1997).
When installed, radiographic equipment is usually set to give a good image
quality at a reasonable radiation dose. This does not necessarily mean that the
system is optimized to a sufficient image quality at the lowest possible dose;
instead it has been shown in this project that there is considerable room for
improvement. To aid in optimization, vendors should include more low-dose
options when installing equipment.
Optimization often reduces image quality. This has to be remembered
when changing settings. There must be a constant discussion with the
radiologists who are responsible for reading the images and it is vital that the
diagnostic yield of a procedure is not jeopardized. Image quality must always
be followed when optimizing a system. Since digital systems are noise-limited,
it is often the increase in noise that limits the attempts at dose reduction,
particularly when simply reducing the number of photons as in a reduction of
the tube charge.

In this project, optimization was carried out in several different ways:


The easiest way of optimizing settings is probably by reducing the tube
charge (mAs). This reduces the number of photons that build the image,
resulting in more pronounced noise. Except for this, the character of the
image is retained because the tube voltage (kV) is unchanged. Using
automatic exposure conditions, this can be accomplished by reducing the
dose request of the detector (papers II, III and IV).
Increased tube voltage (kV) increases the penetration of the beam, which
makes it possible to reduce the mAs and thus reduce dose (paper II). The
increased kV also reduces contrast in the image. However, a solitary in-
crease in kV without compensation for the increased amount of photons
that reach the detector would instead increase dose.
Increased filtration increases the mean keV of the X-ray beam and reduces
the low-energy photons that only cause radiation dose. This is somewhat
similar to an increase in kV. Because of the heavier filtration, the X-ray
tube load is increased (papers IV and V).
Reducing the number of images in a standardized procedure is an obvious
way of reducing the dose (paper II). In a procedure employing different
projections, it is especially beneficial to exclude the projections that deliver
the highest dose.
Exchanging an older detector system for a new, more dose-efficient detec-
tor is very effective. The dose reduction might even be combined with an
increased image quality (paper III).

- 63 -
Hkan Geijer

Removing the grid might also reduce the dose considerably (Martin et al
1999a). This must be balanced with the increased scatter that might occur. It
has a role predominantly in paediatric radiology but can also be used in eg.
chest radiography (Fung and Gilboy 2001).

User habits play a major role in determining the dose from a procedure:
Reduced field size through collimation is very effective in reducing KAP
and effective dose, both in general radiography and in angiography (paper
IV). The use of collimation also increases image quality.
Compression also has the double benefit of reducing dose and increasing
image quality, and this should be used whenever feasible (paper IV).
Ultra-low dose fluoroscopy can often be used as long as the user is able to
increase dose and image quality when needed (papers IV and V). This is
most effective in procedures with high fluoroscopy doses such as inter-
ventional radiology.
Angulation can influence particularly the operator dose in interventional
radiology (paper IV).
Protective screens are very efficient in protecting the operator from scat-
tered radiation (papers IV and V).
Pulsed fluoroscopy is a way to reduce dose in fluoroscopy (den Boer et al
1994; Harrison et al 1998; Wagner et al 2000; Boland et al 2000) (papers
IV and V). Pulsed fluoroscopy should not automatically be regarded as
synonymous with low-dose fluoroscopy since at high frame rates the dose
rate might exceed that of continuous fluoroscopy (Wagner et al 2000).
Ambient lighting should be reduced to get the optimal image quality in
monitor reading (Klein et al 1994; Mertelmeier 1999; Ricke et al 2000) (pa-
per IV). In interventional radiology there is always a balance between
reduced lighting to increase visibility of small objects in the monitor image
and the need of seeing small objects in the room such as balloon catheters
and guidewires that are down to a fraction of a mm in diameter.

Future developments: Launders et al (Launders et al 2001) optimized a digital


selenium-based chest radiography system and found that the optimum balance
between effective dose and image quality was at 100 to 110 kV which is lower
than what is normally used for chest radiography. The results indicate that a
reduction of kV might reduce dose in digital radiography. This has not been
evaluated in the present project.

- 64 -
Discussion

10.7.2 A practical approach to optimization


The following chapter describes a practical approach to the optimization of a
radiographic system. It does not claim to be the definitive answer, however. It
is assumed that the image quality assessment during the optimization is
phantom-based. For best effect, a cooperation of a radiologist, a physicist and
an engineer is recommended.
Make sure the equipment functions properly and that all baseline settings
are correct.
A patient survey before optimization is necessary. The dose can usually be
recorded as KAP values or possibly ESD. For practical purposes, twenty to
thirty patients might be sufficient. To avoid influence of non-standard pa-
tients, inclusion criteria could be set at a weight between 60 and 80 kg for
adults. Alternatively, size correction using the Reference Man concept de-
scribed in chapter 8.1.7 is possible.
Before introducing any changes, baseline settings should be recorded
carefully. This includes equipment settings such as kV, mAs, filtration
and dose request at the detector as well as the geometry such as field size
and focus-to skin distance.
Then, dose and image quality should be assessed. This can be done with
an anthropomorphic phantom and a contrast-detail phantom. Radiation
dose is preferably measured as entrance dose to avoid influence from
varying collimation.
Now it is time to change settings, in the beginning one setting at the time,
later combining settings. All the time, dose and image quality should be
followed using the phantoms.
The resulting images and dose values are then scrutinized. Which change
is most effective in dose reduction with the least detriment in image qual-
ity? This is probably the one that should be instituted. In this stage it is
wise to include more radiologists in the discussion so that more views can
be heard. This discussion is greatly simplified if images of the anthropo-
morphic phantom at different settings are at hand.
Implement the changes. Make sure everything works also when examining
patients.
Finally, a patient survey should be done to evaluate the effect in the clini-
cal population. This should be performed exactly the same way as in the
pre-intervention survey.
Before changing settings, a Monte Carlo simulation can give the answer to the
question of how different changes would influence the effective dose.

Using methods such as the one outlined above, it should be possible to reduce
the radiation exposure to patients and staff in a large amount of procedures.
Still, a constant vigilance on image quality is essential during the optimization.

- 65 -
Hkan Geijer

11 Conclusions
The following conclusions can be drawn from this project:
The scanning method for scoliosis radiography has acceptable image qual-
ity and accuracy in angle measurements in its baseline protocol, but a
high effective dose, about twice that for film. The baseline protocol could
be optimized to a much lower dose with slightly inferior image quality. A
low-dose alternative is to use pulsed fluoroscopy, which delivers one tenth
of the original effective dose.
An amorphous-silicon direct digital flat-panel detector yielded a superior
image quality at a lower dose than both storage phosphor plates and film.
Equivalent image quality to storage phosphor plates was reached at about
one third of the entrance dose.
In percutaneous coronary intervention (PCI), the dose is depending on a
multitude of parameters. By optimizing the fluoroscopy dose rate in PCI, it
was possible to reduce the total KAP for a PCI procedure to two-thirds.
This was assessed in a clinical series of 154 PCI procedures before and
138 after the optimization.

When doing image quality assessments with a phantom, entrance dose is


the preferred dose quantity to measure.
When recording patient doses in a clinical survey, KAP is a valuable
method because it is easy to use and it accounts for changes in collima-
tion as well.
Monte Carlo simulation is a powerful method for estimating the change in
effective dose and thus the cancer risk when changing settings.
The PCXMC computer program for Monte Carlo simulation is easy to use
and seems to be accurate. Using a program such as this makes it much
easier to predict the resulting effective dose of an optimization.
The CDRAD contrast-detail phantom is easy to use and the resulting nu-
merical value of image quality is well correlated to the subjective impres-
sion of image quality.

Optimization of the dose image quality relationship as outlined in this


document is feasible and effective in clinical practice. Such an activity,
however, requires investment in time and effort. Several professional
groups such as radiologists, physicists and engineers must be involved.

- 66 -
Acknowledgements

12 Acknowledgements
I wish to express my sincere gratitude to:
My tutor Jan Persliden, always happy and smiling, always ready to sup-
port me during the dark moments. Jan has guided me into the often in-
comprehensible jungle of medical physics.
My assistant tutor Torbjrn Andersson, always positive this project is
proceeding too fast!.
My co-worker K-W Beckman who has struggled with dose measurements
and Monte Carlo calculations and tried to make sense of the strange re-
sults.
Biomedical engineer Roland Eriksson for help with equipment set-up
and especially for asking those annoying questions like have you thought
of... or why didnt you
Berith Jonsson, RN and Robert Larson, RN for measuring assistance and
loving care of the TLD chips.
Co-author Bert Verdonck at Philips in the Netherlands for help and friend-
ship.
Robert Popek, MD for all the midnight hours doing image quality evalua-
tions.
Lars Norn, MD for help and a sensible approach to image quality.
All radiologists, cardiologists and radiographers at the section for thoracic
radiology for putting up with the rotten fluoroscopic image quality we got
at the end.
Radiographer Lena Evaldsson with colleagues at the section for skeletal
radiology for excellent imaging.
Anders Magnusson, Lennart Bodin, Frida Hgberg and Leif Norn for
statistical assistance.
All the nice and helpful people I have encountered at Philips Medical Sys-
tems in Sweden and the Netherlands.
SYLMAD (Svenska Yngre Lkare Mot Allt Digitalt) for intellectual support,
just like in the best paper I have ever contributed to (Geijer et al 1997).
My brother Mats, MD for countless hours of discussions on digital radiol-
ogy; on the phone, in Vienna, in Ume and elsewhere. He is probably an
even more fervent digital-basher than I am
My neighbours at the dairy farm for giving me ample opportunities to do
something really different for a change.
My wife Mia and my daughter Mrta-Sofie, simply for being there.

This project was supported by grants from:


The research committee of rebro County Council
Swedish Radiation Protection Institute
Swedish Society of Medical Radiology
The Lions Cancer research fund in Sweden
Philips Medical Systems

- 67 -
Hkan Geijer

13 References
PCXMC v 1.4. Users guide. Helsinki, Finland: Finnish Radiation and Nuclear
Safety Authority; 1999.
Alderson SW, Lanzl LH, Rollins M, Spira J. An instrumented phantom system
for analog computation of treatment plans. AJR Am J Roentgenol
1962;87:185-95.
Almn A, Lf M, Mattsson S. Examination technique, image quality, and
patient dose in paediatric radiology. A survey including 19 Swedish
hospitals. Acta Radiol 1996;37:337-42.
Altman DG. Practical statistics for medical research. London: Chapman & Hall;
1991.
Andreo P. Monte Carlo techniques in medical radiation physics. Phys Med Biol
1991;36:861-920.
Antonuk LE, Jee KW, El-Mohri Y, Maolinbay M, Nassif S, Rong X et al.
Strategies to improve the signal and noise performance of active matrix,
flat-panel imagers for diagnostic x-ray applications. Med Phys
2000;27:289-306.
Ardran GM, Coates R, Dickson RA, Dixon-Brown A, Harding FM. Assessment of
scoliosis in children: low dose radiographic technique. Br J Radiol
1980;53:146-7.
Aufrichtig R. Comparison of low contrast detectability between a digital
amorphous silicon and a screen-film based imaging system for thoracic
radiography. Med Phys 1999;26:1349-58.
Aufrichtig R, Xue P. Dose efficiency and low-contrast detectability of an
amorphous silicon x-ray detector for digital radiography. Phys Med Biol
2000;45:2653-69.
Bakalyar DM, Castellani MD, Safian RD. Radiation exposure to patients
undergoing diagnostic and interventional cardiac catheterization
procedures. Cathet Cardiovasc Diagn 1997;42:121-5.
BEIR. Health effects of exposure to low levels of ionizing radiation. BEIR V.
Washington DC, USA: National Academy Press; 1990.
Bernardi G, Padovani R, Morocutti G, Va E, Malisan MR, Rinuncini M et al.
Clinical and technical determinants of the complexity of percutaneous
transluminal coronary angioplasty procedures: analysis in relation to
radiation exposure parameters. Cathet Cardiovasc Interv 2000;51:1-9.
Betsou S, Efstathopoulos EP, Katritsis D, Faulkner K, Panayiotakis G. Patient
radiation doses during cardiac catheterization procedures. Br J Radiol
1998;71:634-9.
den Boer A, de Feyter PJ, Hummel WA, Keane D, Roelandt JRTC. Reduction of
radiation exposure while maintaining high-quality fluoroscopic images
during interventional cardiology using novel X-ray tube technology with
extra beam filtering. Circulation 1994;89:2710-4.
Boland GW, Murphy B, Arellano R, Niklason L, Mueller PR. Dose reduction in
gastrointestinal and genitourinary fluoroscopy: use of grid-controlled
pulsed fluoroscopy. AJR Am J Roentgenol 2000;175:1453-7.

- 68 -
References

Broadhead DA, Chapple CL, Faulkner K, Davies ML, McCallum H. The impact
of cardiology on the collective effective dose in the North of England. Br J
Radiol 1997;70:492-7.
Burgess AE. Comparison of receiver operating characteristic and forced choice
observer performance measurement methods. Med Phys 1995;22:643-55.
Bury RF, Cowen AR, Davies AG, Baker EL, Hawkridge P, Bruijns AJ et al.
Technical report: Initial experiences with an experimental solid-state
universal digital X-ray image detector. Clin Radiol 1998;53:923-8.
Bury RF, Cowen AR, Davies AG, Hawkridge P, Bruijns AJC, von Reth E. Initial
technical and clinical evaluation of a new universal image receptor
system. Eur Radiol 2000;10:1983-7.
Carman DL, Browne RH, Birch JG. Measurement of scoliosis and kyphosis
radiographs. Intraobserver and interobserver variation. J Bone Joint Surg
Am 1990;72-A:328-33.
Chapple CL, Broadhead DA, Faulkner K. A phantom based method for deriving
typical patient doses from measurements of dose-area product on
populations of patients. Br J Radiol 1995;68:1083-6.
Chaussat C, Chabbal J, Ducourant T, Spinnler V, Vieux G, Neyret R. New
CsI/a-Si 17" x 17" X-ray flat-panel detector provides superior detectivity
and immediate direct digital output for general radiography systems. In:
Dobbins JT, Boone JM, editors. Medical Imaging 1998: Physics of Medical
Imaging. SPIE; 1998;3336. p. 45-56.
Chotas HG, Dobbins JT, Ravin CE. Principles of digital radiography with large-
area, electronically readable detectors: a review of the basics. Radiology
1999;210:595-9.
Chotas HG, Ravin CE. Digital chest radiography with a solid-state flat-panel x-
ray detector: contrast-detail evaluation with processed images printed on
film hard copy. Radiology 2001;218:679-82.
Clark AL, Brennan AG, Robertson LJ, McArthur JD. Factors affecting patient
radiation exposure during routine coronary angiography in a tertiary
referral centre. Br J Radiol 2000;73:184-9.
Commission of the European Communities. European guidelines on quality
criteria for diagnostic radiographic images, EUR 16260. Luxembourg:
Office for Official Publications of the European Communities; 1996.
The Council of the European Union. Council Directive 97/43/Euratom. Official
Journal of the European Communities 1997;L (180):22-7.
Cowen AR. Digital X-ray imaging. Measurement Science and Technology
1991;2:691-707.
Cusma JT, Bell MR, Wondrow MA, Taubel JP, Holmes Jr DR. Real-time
measurement of radiation exposure to patients during diagnostic coronary
angiography and percutaneous interventional procedures. J Am Coll
Cardiol 1999;33:427-35.
Dewaele P, Vuylsteke P, Van de Velde S, Schoeters E. Full-leg/full-spine image
stitching: a new and accurate CR-based imaging technique. In: Hanson
KM, editor. Medical Imaging 1999: Image Processing. SPIE; 1999;3661. p.
131-8.
van Eeuwijk AHW, Lobregt S, Gerritsen FA. A novel method for digital X-ray
imaging of the complete spine. In: Troccaz J, editor. Lecture notes in

- 69 -
Hkan Geijer

computer science. Proceedings of CVRMed-MRCAS '97; 1997 March 19-


22; Grenoble, France. Berlin: Springer; 1997;1205. p. 521-30.
Faulkner K, Broadhead DA, Harrison RM. Patient dosimetry measurement
methods. Appl Radiat Isot 1999;50:113-23.
Fearon T, Vucich J, Butler P, McSweeney WJ, Taylor GA, Markle BM et al.
Scoliosis examinations: organ dose and image quality with rare-earth
screen-film systems. AJR Am J Roentgenol 1988;150:359-62.
Floyd Jr CE, Warp RJ, Dobbins JT, Chotas HG, Baydush AH, Vargas-Voracek
R et al. Imaging characteristics of an amorphous silicon flat-panel detector
for digital chest radiography. Radiology 2001;218:683-8.
Frank ED, Stears JG, Gray JE, Winkler NT, Hoffman AD. Use of the
posteroanterior projection: a method of reducing x-ray exposure to specific
radiosensitive organs. Radiol Technol 1983;54:343-7.
Fransson SG, Persliden J. Patient radiation exposure during coronary
angiography and intervention. Acta Radiol 2000;41:142-4.
Fung KK, Gilboy WB. The effect of beam tube potential variation on gonad dose
to patients during chest radiography investigated using high sensitivity
LiF:Mg,Cu,P thermoluminescent dosemeters. Br J Radiol 2001;74:358-67.
Garmer M, Hennigs SP, Jager HJ, Schrick F, van de Loo T, Jacobs A et al.
Digital radiography versus conventional radiography in chest imaging:
diagnostic performance of a large-area silicon flat-panel detector in a
clinical CT-controlled study. AJR Am J Roentgenol 2000;174:75-80.
Geijer M, Jnsson L, Geijer H. New methods for image data compression and
reduction in image archive requirements. AJR Am J Roentgenol
1997;168:901-3.
Geleijns J, Scultze Kool LJ, Zoetelief J, Zweers D, Broerse JJ. Image quality
and dosimetric aspects of chest X-ray examinations: Measurements with
various types of phantoms. Radiat Prot Dosimetry 1993;49:83-8.
Granfors PR, Aufrichtig R. Performance of a 41X41-cm2 amorphous silicon flat
panel x-ray detector for radiographic imaging applications. Med Phys
2000;27:1324-31.
Gray JE, Hoffman AD, Peterson HA. Reduction of radiation exposure during
radiography for scoliosis. J Bone Joint Surg Am 1983;65-A:5-12.
Harrison D, Ricciardello M, Collins L. Evaluation of radiation dose and risk to
the patient from coronary angiography. Aust N Z J Med 1998;28:597-603.
Hart D, Jones DG, Wall BF. Estimation of effective dose in diagnostic radiology
from entrance surface dose and dose-area product measurements.
National Radiological Protection Board publication No. R262. Chilton:
National Radiological Protection Board; 1994.
Hellstrm G, Irstam L, Nachemson A. Reduction of radiation dose in radiologic
examination of patients with scoliosis. Spine 1983;8:28-30.
ICRP. Report of the task group on Reference Man. ICRP Publication 23. Oxford:
Pergamon Press; 1974.
ICRP. Radiation protection. 1977 recommendations of the International
Commission on Radiological Protection. ICRP Publication 26. Oxford:
Pergamon Press; 1977.
ICRP. 1990 recommendations of the International Commission on Radiological
Protection. ICRP Publication 60. Oxford: Pergamon Press; 1991a;21 (1-3).

- 70 -
References

ICRP. Annals of the ICRP. Risks associated with ionising radiations. Oxford:
Pergamon Press; 1991b;22 (1).
ICRP. Radiological protection and safety in medicine. ICRP Publication 73.
Oxford: Pergamon Press; 1996;26 (2).
ICRP. Avoidance of radiation injuries from medical interventional procedures.
ICRP Publication 85. Oxford: Pergamon Press; 2000;30 (2).
ICRU. Phantoms and computational models in therapy, diagnosis and
protection. ICRU Report 48. Bethesda, MD: International Commission on
Radiation Units and Measurements; 1992;48.
ICRU. Medical imaging - the assessment of image quality. ICRU Report 54.
Bethesda, MD: International Commission on Radiation Units and
Measurements; 1996;54.
IPSM. Patient dosimetry techniques in diagnostic radiology. Report No. 53.
York: The Institute of Physical Sciences in Medicine; 1988.
Jansen JTM, Zoetelief J. Computer aided assessment of image quality for
mammography using a contrast detail phantom. Radiat Prot Dosimetry
2000;90:181-4.
Johnson DR, Kyriou J, Morton EJ, Clifton A, Fitzgerald M, Macsweeney E.
Radiation protection in interventional radiology. Clin Radiol 2001;56:99-
106.
Kalmar JA, Jones JP, Merritt CRB. Low-dose radiography of scoliosis in
children. A comparison of methods. Spine 1994;19:818-23.
Katritsis D, Efstathopoulos E, Betsou S, Korovesis S, Faulkner K, Panayiotakis
G et al. Radiation exposure of patients and coronary arteries in the stent
era: A prospective study. Cathet Cardiovasc Interv 2000;51:259-64.
Kengyelics SM, Davies AG, Cowen AR. A comparison of the physical imaging
properties of Fuji ST-V, ST-VA, and ST-VN computed radiography image
plates. Med Phys 1998;25:2163-9.
Kezerashvili M, Rudin S, Bednarek D. Automatic filter placement device for
region of interest (ROI) fluoroscopy. Health Phys 1997;72:141-6.
Klein HM, Stargardt A, Grehl T, Glaser KH, Gnther RW. Einfluss der
Raumbeleuchtung auf die diagnostische Leistung an Bildschirmarbeits-
pltzen. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr
1994;160:168-72.
Koenig TR, Wolff D, Mettler FA, Wagner LK. Skin injuries from fluoroscopically
guided procedures: part 1, characteristics of radiation injury. AJR Am J
Roentgenol 2001a;177:3-11.
Koenig TR, Mettler FA, Wagner LK. Skin injuries from fluoroscopically guided
procedures: part 2, review of 73 cases and recommendations for
minimizing dose delivered to patient. AJR Am J Roentgenol 2001b;177:13-
20.
Kogutt MS, Warren FH, Kalmar JA. Low dose imaging of scoliosis: use of a
computed radiographic imaging system. Pediatr Radiol 1989;20:85-6.
Kohn ML, Gooch Jr AW, Keller WS. Filters for radiation reduction: a
comparison. Radiology 1988;167:255-7.
Kushner DC, Cleveland RH, Herman TE, Zaleske DJ, Ehrlich MG, Correia JA.
Radiation dose reduction in the evaluation of scoliosis: an application of
digital radiography. Radiology 1986;161:175-81.

- 71 -
Hkan Geijer

Labbe MS, Chiu M-Y, Rzeszotarski MS, Bani-Hashemi AR, Wilson DL. The x-
ray fovea, a device for reducing x-ray dose in fluoroscopy. Med Phys
1994;21:471-81.
Laskey WK, Stueve R, Wondrow M, Grill D, Holmes D. Image quality
assessment in contemporary interventional cardiology laboratories: spatial
and low-contrast video resolution. Cathet Cardiovasc Interv 2000;50:257-
63.
Launders JH, Cowen AR, Bury RF, Hawkridge P. Towards image quality, beam
energy and effective dose optimisation in digital thoracic radiography. Eur
Radiol 2001;11:870-5.
Leitz WK, Mnsson LG, Hedberg-Vikstrm BR, Kheddache S. In search of
optimum chest radiography techniques. Br J Radiol 1993;66:314-21.
Levy AR, Goldberg MS, Mayo NE, Hanley JA, Poitras B. Reducing the lifetime
risk of cancer from spinal radiographs among people with adolescent
idopathic scoliosis. Spine 1996;21:1540-7.
Lindskoug BA. Reference man in diagnostic radiology dosimetry. Radiat Prot
Dosimetry 1992;43:111-4.
Ludwig K, Link TM, Fiebich M, Renger B, Diederich S, Oelerich M et al.
Selenium-based digital radiography in the detection of bone lesions:
preliminary experience with experimentally created defects. Radiology
2000;216:220-4.
Manninen H, Terho EO, Wiljasalo M, Wiljasalo S, Soimakallio S. An evaluation
of different imaging chains in clinical chest radiography. Br J Radiol
1984;57:991-5.
Maolinbay M, El-Mohri Y, Antonuk LE, Jee KW, Nassif S, Rong X et al. Additive
noise properties of active matrix flat-panel imagers. Med Phys
2000;27:1841-54.
Marshall NW, Faulkner K. The dependence of the scattered radiation dose to
personnel on technique factors in diagnostic radiology. Br J Radiol
1992;65:44-9.
Marshall NW, Chapple CL, Kotre CJ. Diagnostic reference levels in
interventional radiology. Phys Med Biol 2000;45:3833-46.
Martin CJ, Farquhar F, Stockdale E, MacDonald S. A study of the relationship
between patient dose and size in paediatric radiology. Br J Radiol
1994;67:864-71.
Martin CJ, Hunter S. Reduction of patient doses from barium meal and barium
enema examinations through changes in equipment factors. Br J Radiol
1994;67:1196-205.
Martin CJ, Sutton DG, Sharp PF. Balancing patient dose and image quality.
Appl Radiat Isot 1999a;50:1-19.
Martin CJ, Sharp PF, Sutton DG. Measurement of image quality in diagnostic
radiology. Appl Radiat Isot 1999b;50:21-38.
McParland BJ, Boyd MM. X-ray image intensifier performance and patient
doses for combinations of supplemental beam filters and vascular contrast
agents. Phys Med Biol 2001;46:227-44.
Mertelmeier T. Why and how is soft copy reading possible in clinical practice? J
Digit Imaging 1999;12:3-11.

- 72 -
References

Mooney R, Thomas PS. Dose reduction in a paediatric X-ray department


following optimization of radiographic technique. Br J Radiol 1998;71:852-
60.
Moy JP. Signal-to-noise ratio and spatial resolution in x-ray electronic imagers:
is the MTF a relevant parameter? Med Phys 2000;27:86-93.
Mnsson LG. Evaluation of radiographic procedures. Investigations related to
chest imaging [dissertation]. Gteborg: Gteborg Univ.; 1994.
Nash Jr CL, Gregg EC, Brown RH, Pillai K. Risks of exposure to x-rays in
patients undergoing long-term treatment for scoliosis. J Bone Joint Surg
Am 1979;61-A:371-4.
Neitzel U, Bhm A, Maack I. Comparison of low-contrast detail detectability
with five different conventional and digital radiographic imaging systems.
In: Krupinski EA, editor. Medical Imaging 2000: Image Perception and
Performance. SPIE; 2000;3981. p. 216-23.
Padovani R, Novario R, Bernardi G. Optimisation in coronary angiography and
percutaneous transluminal coronary angioplasty. Radiat Prot Dosimetry
1998;80:303-6.
Persliden J. Application of the Monte Carlo method to diagnostic radiology
[dissertation]. Linkping: Linkping Univ.; 1986;220.
Persliden J, Beckman K-W, Geijer H, Andersson T. Dose-image optimisation in
digital radiology with a direct digital detector. An example applied to pelvic
examinations. Eur Radiol 2001 (in press).
Piraino DW, Davros WJ, Lieber M, Richmond BJ, Schils JP, Recht MP et al.
Direct digital versus conventional film screen radiography of the
musculoskeletal system. J Digit Imaging 1998;11 (3 Suppl 1):172-3.
Pitney MR, Allan RM, Giles RW, McLean D, McCredie M, Randell T et al.
Modifying fluoroscopic views reduces operator radiation exposure during
coronary angioplasty. J Am Coll Cardiol 1994;24:1660-3.
van de Putte S, Verhaegen F, Taeymans Y, Thierens H. Correlation of patient
skin doses in cardiac interventional radiology with dose-area product. Br J
Radiol 2000;73:504-13.
Ricke J, Hanninen EL, Zielinski C, Amthauer H, Stroszczynski C, Liebig T et al.
Shortcomings of low-cost imaging systems for viewing computed
radiographs. Comput Med Imaging Graph 2000;24:25-32.
Rudin S, Bednarek DR, Kezerashvili M, Granger WE, Serghany JE, Guterman
LR et al. Clinical application of region-of-interest techniques to radiologic
imaging. Radiographics 1996;16:895-902.
Shope TB. Radiation-induced skin injuries from fluoroscopy. Radiographics
1996;16:1195-9.
Shrimpton PC, Wall BF, Fisher ES. The tissue-equivalence of the Alderson
Rando anthropomorphic phantom for X-rays of diagnostic qualities. Phys
Med Biol 1981;26:133-9.
Shrimpton PC, Jones DG, Wall BF. The influence of tube filtration and
potential on patient dose during x-ray examinations. Phys Med Biol
1988;33:1205-12.
Siewerdsen JH, Antonuk LE, el-Mohri Y, Yorkston J, Huang W, Cunningham
IA. Signal, noise power spectrum, and detective quantum efficiency of
indirect-detection flat-panel imagers for diagnostic radiology. Med Phys
1998;25:614-28.

- 73 -
Hkan Geijer

Spahn M, Strotzer M, Vlk M, Bhm S, Geiger B, Hahm G et al. Digital


radiography with a large-area, amorphous-silicon, flat-panel X-ray
detector system. Invest Radiol 2000;35:260-6.
Stringer DA, Cairns RA, Poskitt KJ, Bray H, Milner R, Kennedy B. Comparison
of stimulable phosphor technology and conventional screen-film
technology in pediatric scoliosis. Pediatr Radiol 1994;24:1-5.
Strotzer M, Gmeinwieser J, Spahn M, Vlk M, Frund R, Seitz J et al.
Amorphous silicon, flat-panel, X-ray detector versus screen-film
radiography: effect of dose reduction on the detectability of cortical bone
defects and fractures. Invest Radiol 1998a;33:33-8.
Strotzer M, Gmeinwieser JK, Vlk M, Frund R, Seitz J, Feuerbach S. Detection
of simulated chest lesions with normal and reduced radiation dose:
comparison of conventional screen-film radiography and a flat-panel X-ray
detector based on amorphous silicon. Invest Radiol 1998b;33:98-103.
Strotzer M, Vlk M, Wild T, von Landenberg P, Feuerbach S. Simulated bone
erosions in a hand phantom: detection with conventional screen-film
technology versus cesium iodide-amorphous silicon flat-panel detector.
Radiology 2000;215:512-5.
Svik E, Klw N-E, Hellesnes J, Lykke J. Radiation-induced skin injury after
percutaneous transluminal coronary angioplasty. Acta Radiol
1996;37:305-6.
Tapiovaara M, Lakkisto M, Servomaa A. PCXMC - A PC-based Monte Carlo
program for calculating patient doses in medical X-ray examinations.
Version 1.3. Helsinki, Finland: Finnish Centre for Radiation and Nuclear
Safety; 1997;STUK-A139.
Thijssen MAO, Thijssen HOM, Merx JL, Lindeijer JM, Bijkerk KR. A definition
of image quality: the image quality figure. In: Moores BM, Wall BF, Eriskat
H, Schibilla H, editors. BIR Report 20: Optimization of image quality and
patient exposure in diagnostic radiology. London: The British Institute of
Radiology; 1989. p. 29-34.
Thijssen MAO, Bijkerk KR, van der Burght RJM. Manual CDRAD-phantom
type 2.0. Project Quality Assurance in Radiology, Section Clinical Physics,
Department of Radiology. St. Radboud, The Netherlands: University
Hospital Nijmegen; 1998.
Tingberg A. Quantifying the quality of medical X-ray images [dissertation].
Malm: Lund Univ.; 2000.
Tsuno MM, Shu GJ. Posteroanterior versus anteroposterior lumbar spine
radiology. J Manipulative Physiol Ther 1990;13:144-51.
Va E, Gonzalez L, Fernandez JM, Guibelalde E. Patient dose values in
interventional radiology. Br J Radiol 1995a;68:1215-20.
Va E, Guibelalde E, Morillo A, Alvarez-Pedrosa CS, Fernandez JM.
Evaluation of the European image quality criteria for chest examinations.
Br J Radiol 1995b;68:1349-55.
Va E, Arranz L, Sastre JM, Moro C, Ledo A, Garate MT et al. Dosimetric and
radiation protection considerations based on some cases of patient skin
injuries in interventional cardiology. Br J Radiol 1998;71:510-6.
Va E, Gonzalez L, Ten JI, Fernandez JM, Guibelalde E, Macaya C. Skin dose
and dose-area product values for interventional cardiology procedures. Br
J Radiol 2001;74:48-55.

- 74 -
References

Verdonck B, Nijlunsing R, Melman N, Geijer H. Image quality and X-ray dose


for translation reconstruction overview imaging of the spine, colon and
legs. In: Lemke HU, Vannier MW, Inamura K, Farman AG, Doi K, editors.
CARS 2001. Proceedings of Computed Assisted Radiology and Surgery;
2001 June 27-30; Berlin, Germany. Amsterdam: Elsevier Science; 2001.
p. 500-5.
Vlk M, Strotzer M, Gmeinwieser J, Alexander J, Frund R, Seitz J et al. Flat-
panel X-ray detector using amorphous silicon technology. Reduced
radiation dose for the detection of foreign bodies. Invest Radiol
1997;32:373-7.
Vlk M, Strotzer M, Holzknecht N, Manke C, Lenhart M, Gmeinwieser J et al.
Digital radiography of the skeleton using a large-area detector based on
amorphous silicon technology: image quality and potential for dose
reduction in comparison with screen-film radiography. Clin Radiol
2000;55:615-21.
Wagner LK, McNeese MD, Marx MV, Siegel EL. Severe skin reactions from
interventional fluoroscopy: case report and review of the literature.
Radiology 1999;213:773-6.
Wagner LK, Archer BR, Cohen AM. Management of patient skin dose in
fluoroscopically guided interventional procedures. J Vasc Interv Radiol
2000;11:25-33.
Widmark A, Fosmark H, Einarsson G, Gundtoft P, Hjardemaal O, Leitz W et al.
Guidance levels in the Nordic countries: a preliminary report for selected
interventional procedures. Radiat Prot Dosimetry 2001;94:133-5.
Wilde P, Pitcher EM, Slack K. Radiation hazards for the patient in cardiological
procedures. Heart 2001;85:127-30.
Yaffe MJ, Rowlands JA. X-ray detectors for digital radiography. Phys Med Biol
1997;42:1-39.
Zorzetto M, Bernardi G, Morocutti G, Fontanelli A. Radiation exposure to
patients and operators during diagnostic catheterization and coronary
angioplasty. Cathet Cardiovasc Diagn 1997;40:348-51.

- 75 -
Hkan Geijer

- 76 -

You might also like