Professional Documents
Culture Documents
Pe d i a t r i c I m a g i n g • C l i n i c a l O b s e r v a t i o n s
Neonatal Computed
Radiography
Downloaded from www.ajronline.org by 115.178.236.142 on 03/22/19 from IP address 115.178.236.142. Copyright ARRS. For personal use only; all rights reserved
A B
4,500
4,000
Lung
3,500 Lowest brightness
Output Code Value
3,000
2,500 Middle brightness
2,000
1,500 Highest brightness
1,000
Bone
500
0
1,000 1,200 1,400 1,600 1,800 2,000 2,200 2,400 2,600 2,800 3,000
Input Code Value
C D
Fig. 1—Normal neonatal chest radiographs illustrate response curve to brightness adjustments for neonatal chest computed radiography. Lungs appear progressively lighter
from lowest brightness to highest brightness.
A, Low-brightness image.
B, Reference T-MAT G (Eastman Kodak Company) image.
C, High-brightness image.
D, Graph shows adjustment of characteristic curve from lowest to highest brightness. At fixed input code value, lowest brightness images have higher or more lung density
output and highest brightness images have lower or more bone density output.
Materials and Methods Imaging Processing of the response curve. Low-detail contrast images
Image Acquisition The raw image data were processed with a proto- appear grayer, and high-detail contrast images are
This study received the approval of our institu- type image-processing algorithm [8]. Three variables more black and white (Fig. 2). This control is anal-
tional review board. All image data were handled were altered: brightness, detail contrast, and latitude. A ogous to the window in CT in which shallow lung
according to the Health Insurance Portability and control image for each patient was acquired to simulate window contrast minimizes the contrast difference
Accountability Act (HIPAA). Images were ob- the characteristic response curve of T-MAT G film between water and soft tissue and steep mediasti-
tained with a Kodak 400 CR system (Eastman (Eastman Kodak Company). This screen-film combi- num window contrast accentuates the difference
Kodak Company). Five chest radiographs of nation is common in pediatric chest radiography. between water and soft tissue.
healthy neonates and five chest radiographs of ne- Brightness is increased on a CR image through Latitude adjustment on a CR image changes the
onates with clearly discernible pneumothorax were a shift in the response curve to the right (Fig. 1). overall contrast of an image without affecting local
selected. Raw image data were collected at a qual- This control is analogous to the level in CT image contrast or resolution. For a wide-latitude image, one
ity-control workstation with all image processing display in which the lungs are dark at a mediastinal compresses the large-area, low-frequency data of the
turned off. These data were transferred to another level and bright at a lung level. histogram information and subtracts that from the
workstation for removal of all patient identification Detail contrast is increased on a CR image histogram of the image. The overall appearance of
data and then were sent to an analysis workstation. through an increase in the slope of the linear portion the image is grayer yet retains local contrast. For ex-
A B
Fig. 2—Normal neonatal chest radiographs show detail contrast adjustment of response
4,500
curve for neonatal chest computed radiograph. See reference image in Figure 1B for
4,000 comparison. Difference in contrast between air in stomach and ribs is evident.
Lung
Lowest contrast = 2.5
3,500 A, Low-detail contrast image appears gray.
Output Code Value
4,500
4,000
Lung
3,500
Output Code Value
Latitude = 0.56
3,000
2,500
2,000
1,500
1,000 Latitude = 1.46
Bone 500
0
0 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000
Input Code Value
A B
Fig. 3—Normal neonatal chest radiograph with latitude adjustment of response curve for neonatal chest computed radiograph. See reference image in Figure 1B for
comparison. Difference in free-in-air exposure and soft tissue around humerus is evident.
A, Wide-latitude image appears gray.
B, Graph shows adjustment of characteristic curve. As latitude increases, steepness of response curve decreases. This finding applies to low-spatial-frequency data only.
ample, the contrast between a rib and adjacent lung tures such as lung markings and bone detail and min- contrast, and latitude for a neonate with pneumotho-
is maintained, but the contrast between the free-in- imizes differences such as density difference be- rax are presented in Figure 4.
air exposure and soft tissues around the humerus is tween lungs. Examples of the effect of altering the Each case was processed with five brightness
decreased (Fig. 3). This manipulation highlights fea- image-processing parameters of brightness, detail levels, four detail contrast levels, and seven latitude
A B C
D E F
Fig. 4—Neonatal chest radiographs with pneumothorax. Images show effects of brightness, detail contrast, and latitude adjustment on neonatal chest radiographs.
A, Reference image.
B, Low-brightness image.
C, High-brightness image.
D, Low-contrast image.
E, High-contrast image.
F, Wide-latitude image.
levels. One hundred ten combinations of 140 possi- middle brightness level was chosen as the level that justments were achieved by reducing the low-fre-
ble combinations of each patient’s image were would yield an optical density of approximately 1.65 quency contrast at set increments while maintaining
made. The processing parameter matrix was not on a T-MAT G film image of the lung region. Bright- the detail contrast, achieved through a signal-equal-
symmetric, and the narrower-latitude and lower- ness adjustments were achieved by shifting the tone ization filtering step.
detail contrast images were not acquired because scale along the input axis in increments of 75 and The control image-processing parameters were
observation [8] had shown the range of lowest lati- 150 code values above and below the middle bright- chosen as the image at middle brightness (one that
tude and contrast performed the worst. With this ness code value for each image to yield lowest, would yield an optical density of approximately
criterion, 30 image combinations per patient were lower, middle, high, and highest brightness settings. 1.65 on T-MAT G film in the lung region), low-de-
eliminated without alteration of the results, allow- At the reference detail contrast setting, that adjust- tail contrast (T-MAT G film contrast of 3.1), and
ing reduction of computational time and enhance- ment would yield optical densities in the lung region narrow latitude (one that would maintain the tonal
ment of observer participation. of, from low to high, 2.25, 1.95, 1.65, 1.35, and 1.15. characteristics of low-frequency data for a detail
Combinations were achieved by adjustment of Detail contrast adjustments were achieved by pivot- contrast level of 3.1). These parameters would sim-
the tone scale with a fixed shape that approximately ing the tone scale at set increments around the input ulate on soft-copy display the hard-copy screen-
matched the tone scale curve of T-MAT G film. The value corresponding to a density of 1.0. Latitude ad- film appearance of neonatal chest radiographs.
Image Viewing Results were initially displayed from fixed, preset im-
One hundred ten images were acquired for each The regression results showed that bright- age-processing algorithms. In our study, the
of 10 sets of neonatal CR images for a total of 1,100 ness (p < 0.01) was the single most impor- image-processing parameters were the tested
images. These images were randomized and pre- tant factor in determining user imaging and variables. In three of the previous studies
sented in 10 sessions of 110 images. The images diagnostic preferences (Table 1). Also sig- [12–14], it was explicitly stated that radiolo-
Downloaded from www.ajronline.org by 115.178.236.142 on 03/22/19 from IP address 115.178.236.142. Copyright ARRS. For personal use only; all rights reserved
were viewed with a stand-alone PC workstation on nificant in order of contribution to the vari- gists were allowed to adjust window and level
a high-resolution (2,000 × 2,500) cathode ray tube ance were observer, detail contrast, latitude, but not spatial filtering. In one study [14], the
monitor (model DR 110, Data Ray). The monitor and patient (Table 1). The presence or ab- radiologists manipulated window and level
was calibrated, and lighting was controlled to elim- sence of pneumothorax was not statistically settings 90% of the time. In only one study [8],
inate glare on the screen. The workstation was in a significant (p = 0.14). with adult subjects, were preferred soft-copy
quiet room for elimination of extraneous noise. There was a peak cell-rating difference of display settings evaluated.
Each test image was presented along with the three points between the user-preference pre- Our study of neonatal chest CR images
control processed image. The observer toggled be- dicted best cell rating at the lowest brightness showed that pediatric radiologists find im-
tween the test image and the control image and con- (–1.7) and the user-preference predicted best proved image and diagnostic quality of soft-
trolled the toggle rate and amount of observation cell rating at the highest brightness (1.3). At copy display with the brightest image setting
time before rating. Three pediatric radiologists with the lowest brightness, all scores were signifi- and with high-detail contrast and narrow to
certificates of added qualification in pediatric radi- cantly less than 0, indicating that the images middle latitude processing settings. Adjusting
ology participated in the image review. These radi- at this brightness level, regardless of detail the CR image display resulted in a higher rat-
ologists were familiar with soft-copy interpretation. contrast and latitude, were less desirable than ing by the radiologists compared with the ref-
Each radiologist rated each image on a nine-point the reference image in image and diagnostic erence screen-film radiographic appearance.
viewing scale relative to the control image as in pre- quality (Table 2). At the middle brightness The CR viewing preference was different from
viously published adult work [8]: 4, image quality level, which included the reference image, the the screen-film radiographic appearance of the
markedly better, diagnosis likely altered; 3, image predicted ratings were near 0 (Table 3). The characteristic response curve of T-MAT G
quality clearly better, diagnosis might be altered; 2, ratings at the highest brightness and highest film, which is middle brightness, narrow lati-
image quality somewhat better, diagnosis should be detail contrast and narrow latitude had the tude, and less detail contrast. Wide-latitude im-
the same; 1, image quality slightly better, diagnosis highest predicted score (Table 4). At none of ages were not preferred for routine viewing.
will be the same; 0, no difference; –1, image quality the image brightness settings was there a sta- The parameters that were varied in this ex-
slightly worse, diagnosis will be the same; –2, im- tistically significant difference between the periment (brightness, detail contrast, and lat-
age quality somewhat worse, diagnosis should be ratings for healthy patients and those for pa- itude) are the building blocks of image qual-
the same; –3, image quality clearly worse, diagno- tients with pneumothorax. ity. As such, the results can be generalized for
sis might be altered; –4, image quality markedly use with most CR systems, provided the inter-
worse, diagnosis likely altered. Discussion face with image processing has the required
Despite the presence of CR in pediatric im- flexibility. Although the particular image pro-
Statistical Analysis aging for nearly 20 years [9], to our knowl- cessing (rendering) parameters are different
A regression model with a second-order polyno- edge no scientific methodology has been de- among CR systems, the functionality of im-
mial was used to analyze the data. The factors of veloped for addressing specific soft-copy age processing can be related to these three
brightness, latitude, and detail contrast were treated viewing preferences. Years ago, a hard-copy fundamentals of image quality.
as continuous. A logarithmic transform of the lati- CR film included two images, one imitating a The results of this study can be applied to
tude was used to improve conformance to the poly- screen-film radiograph and a second that en- soft-copy and hard-copy interpretation, pro-
nomial fit. Brightness, detail contrast, and log-lati- hanced edge structures, such as catheters [8, vided the image-processing parameters are
tude levels were centered to reduce the correlation 9]. Currently, one hard-copy CR film melds adapted to compensate image appearance for
between predictor values, thus clarifying signifi- the screen-film appearance and edge en- the dynamic range, gray-scale resolution, and
cance testing. In addition, health condition (pres- hancement. Our findings clearly indicate that spatial resolution differences among printers
ence or absence of pneumothorax) was included as soft-copy viewing preferences in neonatal and soft displays. The concept of the building
a discrete and fixed effect. Observer and patient ef- chest CR are different from the established blocks of image quality has been detailed [17].
fects were treated as discrete and random, and the hard-copy screen-film radiographic prefer- In a previous study [8] of posteroanterior
patient effect was considered nested within the ences [8]. Our findings establish baseline chest radiographs of adults, the radiologists
health condition. soft-copy settings for future use in image dis- preferred a wide-latitude (highly equalized)
The peak soft-copy viewing preference range of play and dose-reduction research. image with less detail contrast, different from
detail contrast and latitude cell at each brightness More radiology departments are shifting to the user preferences we found. One explana-
level was determined. The best-rated cell at each soft-copy display and eliminating film entirely. tion may be that adult chest images have a
brightness level was identified. The 95% CI was de- Much work has been done to meet the chal- broader histogram range than chest images of
termined for the difference between the predicted lenge of PACS and soft-copy display [10]. Re- neonates. This difference is attributable to the
mean of the cell with the best rating and the pre- search has focused on comparing soft-copy size difference between neonates and adults.
dicted mean of the cell being checked. All cells display with hard-copy film [11], resolution of The anatomic structures of neonates are less
identified as not significantly different from the the monitor [11], CR with other digital radio- attenuating than are those of adults. The typi-
best rating cell were included in the peak prefer- graphic systems [12, 13], and monitor lumi- cal peak kilovoltage used for neonatal chest
ence range for each brightness level. nance [14–16]. In these studies, the images radiographs is 60–70 kVp, whereas that for
TABLE 1: Regression Model Results This finding implies that there is no need for
Source of Variation Degrees of Freedom F p routine alteration of image display for routine
Brightness 1 2,734.44 < 0.01a evaluation for pneumothorax. This study, how-
ever, did not test ability to diagnose subtle cases
Observer 2 235.68 < 0.01a
of pneumothorax. A previous study of neonatal
Detail contrast 1 99.12 < 0.01a
Downloaded from www.ajronline.org by 115.178.236.142 on 03/22/19 from IP address 115.178.236.142. Copyright ARRS. For personal use only; all rights reserved
References 7. Roehrig H, Krupinski EA, Hulett R. Reduction of chest lesions by using soft-copy reading: compari-
1. Curry TS, Dowdey JE, Murry RC. Photographic patient exposure in pediatric radiology. Acad Radiol son of an amorphous silicon flat-panel detector sys-
characteristics of x-ray film. In: Christensen’s phys- 1997; 4:547–557 tem and a storage-phosphor system. Radiology
ics of diagnostic radiology. Malvern, PA: Lea & Fe- 8. Flynn M, Couwenhoven M, Eyler W, et al. Optimal 2002; 224:242–246
biger, 1990:148–164 display processing for digital radiography. In: SPIE 14. Goo JM, Choi J, Im J, et al. Effect of monitor lumi-
Downloaded from www.ajronline.org by 115.178.236.142 on 03/22/19 from IP address 115.178.236.142. Copyright ARRS. For personal use only; all rights reserved
2. Hillman BJ, Fajardo LL. Clinical assessment of medical imaging 2001: visualization, display and nance and ambient light on observer performance in
phosphor-plate computed radiography: equip- image guide procedures. Piscataway, NJ: IEEE, soft-copy reading of digital chest radiographs. Ra-
ment, strategy, and methods. J Digit Imaging 2001:298–305 diology 2004; 232:762–766
1989; 2:220–227 9. Dietrich RB, Boechat MI, Huang HK. Experi- 15. Ikeda M, Ishigaki T, Shimamoto K, et al. Influence
3. Cowen AR, Workman A, Price JS. Physical aspects ence with phosphor imaging plates: clinical ex- of monitor luminance change on observer perfor-
of photostimulable phosphor computed radiogra- perience in pediatric radiology. Proc SPIE 1989; mance for detection of abnormalities depicted on
phy. Br J Radiol 1993; 66:332–345 1091:242–244 chest radiographs. Invest Radiol 2003; 38:57–63
4. Don S. Radiosensitivity of children: potential for 10. Arenson RL, Chakroborty DV, Seshadri SB, Kun- 16. Ishihara S, Shimamoto K, Ikeda M, et al. CRT di-
overexposure in CR and DR and magnitude of doses del HL. The digital imaging workstation. Radiology agnosis of pulmonary disease: influence of mon-
in ordinary radiographic examinations. Pediatr Ra- 1990; 176:303–315 itor brightness and room illuminance on observer
diol 2004; 34[suppl 3]:S167–S172 11. Otto D, Bernhardt TM, Rapp-Bernhardt U, Lud- performance. Comput Med Imaging Graph 2002;
5. Don S, Cohen MD, Kruger RA, et al. Volume de- wig K, Liehr UB, Dohring W. Subtle pulmonary 26:181–185
tection threshold: quantitative comparison of abnormalities: detection on monitors with vary- 17. Couwenhoven M, Senn R, Foos D. Enhancement
computed radiography and screen-film radiogra- ing spatial resolutions and maximum luminance method that provides direct and independent con-
phy in detection of pneumothoraces in an animal levels compared with detection on storage phos- trol of fundamental attribute of image quality for ra-
model that simulates the neonate. Radiology phor radiographic hard copies. Radiology 1998; diographic imagery. In: SPIE Medical Imaging
1995; 194:727–730 207:237–242 2003: visualization, display and image guide pro-
6. Don S, Hildebolt C, Sharp T, et al. Comparison 12. Goo JM, Im J, Kim JH, et al. Digital chest radi- cedures. Piscataway, NJ: IEEE, 2004:474–481
of computed radiography with film-screen radi- ography with a selenium-based flat-panel detec- 18. Goo HW, Kim H, Song K, et al. Using edge en-
ography in detecting pulmonary edema in a rabbit tor versus a storage phosphor system. AJR 2000; hancement to identify subtle findings on soft-
model simulating the neonate. Radiology 1999; 175:1013–1018 copy neonatal chest radiographs. AJR 2001;
213:455–460 13. Goo JM, Im J, Lee HJ, et al. Detection of simulated 177:437–440