British Journal of Radiology (2005) 78, 993-996
© 2005 British Institute of Radiology
doi: 10.1259/bjr/30262671
Influence of film digitization on radiological interpretation
T Pudas, MD1,
L Korsoff, MD1,
T Kallio, MD, PhD1,
M Uhari, MD, PhD2 and
A Alanen, MD, PhD1
1 Turku University Central Hospital, PL52, 20521 Turku, Finland and 2 University of Oulu, Department of Pediatrics, 90014 Oulu, Finland
Correspondence: Tomi Pudas, Kiinamyllynkatu 4-8 20520 Turku, Finland
 |
Abstract
|
|---|
Our objective was to evaluate the influence of changing from analogue to digital imaging on interobserver and intraobserver image interpretation. Three radiologists interpreted 96 three image series of occipitomental radiographs of paranasal sinuses from the films and from the corresponding digitized images from the screen. Images were classified according to degree of abnormality as either normal, with mucosal thickening of less than 5 mm, with mucosal thickening of 5 mm or more, total opacity, air–fluid level or polyp or cyst of maxillary sinuses. In the present study we found that there were more differences between two radiologist's interpretations with a single method than in a single radiologist's interpretations between the methods, although radiologists interpreted fewer pathological findings from the digitized images than from the corresponding films. Our data show that the results of image interpretation are preferentially dependent on the reader rather than on the method of reading.
|
Introduction
|
|---|
In recent years digital imaging has generally superseded conventional analogue imaging in radiology, which has led to a profound change in the interpretation of radiological images. The images are read from a work station and the use of conventional hardcopies has decreased markedly. Analogue images are interpreted against a light board which enables the radiologist to view several images concomitantly, and this is especially valuable when old and new studies are compared. Digital images are read on a screen at a work station in succession or simultaneously, depending on the size and capacity of the screen, the software and the amount of screens. Digital images can be post-processed at the work station and the diagnostic output may thus be increased.
Many studies concerning the quality and diagnostic reliability of digital images vs analogue images have been made [1–4]. The optimal matrix (or pixel) size and the amount of grey scales for reliable diagnostics are now known. However, the influence of the change in working manners of the readers has only seldom been evaluated.
In the present study we compared interpretation of digitized images with original analogue images. We recorded the interobserver difference between the three readers, with the following question in mind: are there significant differences when radiologists read analogue images or digitized images? A specific goal was to evaluate the change in the diagnostic accuracy when digitized rather than analogue images were examined. Plain radiography of the paranasal sinuses was chosen as a model because the image is of a suitable size for digitizing and because a large amount of material was available from another study dealing with viral sinusitis [5].
|
Materials and methods
|
|---|
The study was performed at the Turku University Central Hospital (TUCH). Images from 105 young adults were used. Plain radiographs of the paranasal sinuses were taken in the occipitomental view on the first day, 6 days later and 21 days later for the patients who had the common cold. For digitizing, a code number was given to each individual radiograph and the images were interpreted independently by three radiologists who were blinded to the clinical patient data.
The plain radiographs were digitized with a Pinja digitizer (Finnelpro, Tampere, Finland) to a resolution of 4000 x 4000 pixels and a dynamic range of 16 bits. The digitized images were saved in the digital image archive of TUCH. The images were then downloaded to a work station (AutoRad 3.0; Cemax-Icon, Fremont, CA) for interpretation at the leisure of the radiologists. The images could be windowed and zoomed. The series of three consecutive images were present at the same time as the interpretation session to allow comparison of the images over time. The analogue films and the digitized images were interpreted at different sessions 2 years apart.
Because of technical reasons, the images of nine patients were not interpreted by all the radiologists. These images were excluded from the analysis (Tables 1–3
).
View this table:
[in this window]
[in a new window]
|
Table 1. Categorical findings of the maxillary sinuses by Radiologist A. The interpretation after image digitization in certain category of film image findings
|
|
View this table:
[in this window]
[in a new window]
|
Table 2. Categorical findings of the maxillary sinuses by Radiologist B. The interpretation after image digitization in certain category of film image findings
|
|
View this table:
[in this window]
[in a new window]
|
Table 3. Categorical findings of the maxillary sinuses by Radiologist C. The interpretation after image digitization in certain category of film image findings
|
|
The findings of the maxillary sinuses were categorized as: (1) normal; (2) mucosal thickening of 5 mm or less; (3) mucosal thickening of more than 5 mm; (4) total opacity; (5) air–fluid level; and (6) cyst or polyp. Only one categorical number per image was allowed.
Statistical analyses
The weighted kappa statistic was used to evaluate the level of agreement between radiologist and methods, and the results are shown in Tables 4 and 5. For analysis of weighted kappa we recorded findings according to severity, as follows: (1) normal; (2) mucosal thickening of 5 mm or less; (3) mucosal thickening of more than 5 mm; (4) air–fluid level; (5) total opacity. Polyp or cyst was not included as a category when weighted kappa analysis was performed. The results were analysed with SPSS-software (SPSS Inc., IL). The variation of the each categorical finding of the maxillary sinuses is presented in Tables 1–3
. Only the patients who had all three image series interpreted by all three radiologists were analysed. A total of 96 patients met the criteria for inclusion.
View this table:
[in this window]
[in a new window]
|
Table 4. Intrapersonal kappa statistics between the observers: analogue vs digitized images of paranasal sinuses on day one of the common cold. (95% confidence interval)
|
|
View this table:
[in this window]
[in a new window]
|
Table 5. Kappa statistics between analogue and digitized images. Kappa statistics between analogue and digitized images interpretated by radiologists from the radiography of the paranasal sinuses on day one of the common cold. The interpretation of kappa values is presented in Table 6
|
|
|
Results
|
|---|
The correlation between the interpretations of the maxillary sinus findings on analogue films on the first day of the common cold and the corresponding digitized images were analysed (Table 5). The correlation for all radiologists was moderate, although radiologist C had relatively lower agreement (kappa values were 0.529 for radiologist A, 0.568 for radiologist B and 0.424 for radiologist C).
Table 4 shows the kappa values of the interpretation between the different radiologists. The interpretations made by radiologists A and B agreed moderately regarding the interpretation of films (kappa 0.541) and digitized images (kappa 0.518). Radiologists C and A had a fair agreement for their interpretation of films (kappa 0.357) and digitized images (kappa 0.269). Between radiologists C and B the agreement was fair by both methods, while the kappa value was slightly higher for the analogue films (kappa 0.433 for film and kappa 0.253 to digitized images).
Categorical findings of 576 maxillary sinuses
The findings in the images of both maxillary sinuses of 96 patients (occipitomental view) at days 1, 7 and 21 are shown in Tables 1–3. The categorical findings from all 576 interpreted sinuses were counted from both film and digitized image readings and the results are presented in parallel.
Normal findings
Radiologists A and B interpreted films and digitized images almost in parallel. Radiologist A categorized 426 maxillary sinuses as normal on films, and 392 of those were interpreted as normal after digitization. The corresponding results for radiologist B were 397 and 375 and for radiologist C 313 and 179. Radiologist C clearly identified fewer normal findings than the others. Radiologist C categorized 79 and 49 sinuses as having mucosal thickening after digitization (Table 3). Radiologists A and B did not detect any air–fluid levels in the digitized images that they had categorized as normal on films, whereas radiologist C detected three such instances.
Radiologist A categorized 437 and radiologist B 446 digitized images as normal, which is a bit more than their respective findings on films. Radiologist C categorized 212 maxillary sinuses as normal and 318 sinuses (158 and 160) as having mucosal thickening in the digitized image group.
There was a trend for A and B to categorize the digitized images more often as normal than for radiologist C.
Mucosal thickening
Radiologist A interpreted the mucosal thickening of 47 maxillary sinuses as being 1–5 mm and 55 as more than 5 mm on film. After digitizing, 30 of the 47 sinuses with 1–5 mm mucosal thickening and 6 of the 55 with a thickening of more than 5 mm were interpreted as normal (Table 1). The corresponding results for radiologist B was 86–49 and 42–13 (Table 2) and for radiologist C 119–27 and 103–4 (Table 3). Radiologist C interpreted clearly more images as exhibiting mucosal thickening, especially after digitization, than the others.
Air–fluid level
Radiologist A detected 11 maxillary sinuses as containing an air–fluid level on films, B detected 19 and C 15. Radiologist A interpreted five (45%) of these as normal and two (18%) as mucosal thickening of more than 5 mm after digitization (Table 1). Radiologist B interpreted six of 19 (32%) as normal after digitization. Radiologist C had two of 15 (13%) interpreted as normal and two (13%) as mucosal thickening after digitization.
Total opacity
Radiologist A detected 14 maxillary sinuses with total opacity, B detected 10 and C 3. Radiologist A interpreted six of 14 as having mucosal thickening, polyp/cyst or air–fluid level in digitized images. No images were interpreted as normal after digitizing (Table 1). Radiologist B had similar results (Table 2), while radiologist C detected only three total opacities which remained unchanged after digitization. On the other hand, radiologist C found 13 total opacities in the digitized images; 7 were categorized as mucosal thickening when plain films were inspected.
Polyp or cyst
A similar number of maxillary sinuses were interpreted as containing polyp or cyst: radiologist A found 23, B 22 and C 23. Of these, 17, 13 and 15 sinuses, respectively, had no change of interpretation after digitization. All radiologists interpreted one (not the same) as an air–fluid level. Some were detected as mucosal thickening, but only radiologist C categorized none as normal.
|
Discussion
|
|---|
In this study we analysed the influence of digitization on radiological decision making. Digitization was made with a Pinja digitizer. Three experienced radiologists (LK, TK and AA) analysed the images for mucosal thickening, air–fluid level, total opacity, polyps or cysts from both maxillary sinuses and indicated their findings with the numbers 1–6. The radiographic findings were not assessed by other more sophisticated methods (CT or MRI). The radiologists were allowed to allocate only one categorical number to each maxillary sinus. For example a maxillary sinus with mucosal thickening and an air–fluid level was categorized by the most important finding, in this particular case as an air–fluid level. The kappa values were higher for the same radiologist interpreting films and digitized images than for two different radiologist analysing images by either method interindividually. This suggests that the results of image interpretation are preferentially dependent on the reader rather than on the method of image reading. Experienced radiologists can estimate the findings independent of the method used. Expensive digitization methods and displays are not so important for image analysis, but every day work may be easier if good quality monitors are available. Some special cases, like diffuse parenchymal diseases of the lungs and pneumothorax, need high resolution displays [6, 7].
The interpretations of normal findings by two young radiologists (A and B) stood in parallel when films and digitized images were reviewed. There were slightly fewer findings of mild mucosal thickening after digitization. Radiologist C identified more often mucosal thickening, while the interpretation of an air–fluid level was very similar by both methods. On closer inspection, digitization by scanning the films seems, infact, to lose pathological findings rather than redundancy to some extent. This needs more investigation with appropriate control imaging methods. The scale of a digitized image on the screen is different than the size of a radiograph, which might influence the estimation of mucosal thickening. Detection of an air–fluid level or of total opacity should not be affected by image size. At the time of the study, work stations were only being introduced into clinical practice. If the study were done now, direct digital imaging facilities would be chosen and direct digital images on our diagnostic monitors would be compared with hard copies printed from those monitors. In our opinion, when a radiological unit changes from analogue to digital radiography, it is important that valid working procedures and quality systems are used from the very beginning.
Received for publication March 2, 2005.
Revision received April 27, 2005.
Accepted for publication May 13, 2005.
|
References
|
|---|
- Brill PW, Winchester P. Hard- versus soft-copy readings of computer radiographs. Radiology 1998;208:270–1.[Medline]
- Brill PW, Winchester P, Cahill P, Lesser M, Durfee SM, Giess CS, et al. Computed radiography in neonatal and pediatric intensive care units: a comparison of 2.5 K x 2 K soft-copy images vs digital hard-copy film. Pediatr Radiol 1996;26:333–6.[CrossRef][Medline]
- Kundel HL, Polansky M, Dalinka MK, Choplin RH, Gefter WB, Kneelend JB, et al. Reliability of soft-copy versus hard-copy interpretation of emergency department radiographs: a prototype study. AJR Am J Roentgenol 2001;177:525–8.[Abstract/Free Full Text]
- Kundel HL, Gefter W, Aronchick J, Miller W Jr, Hatabu H, Whitfill CH, et al. Accuracy of bedside chest hard-copy screen-film versus hard- and soft-copy computed radiographs in a medical intensive care unit: receiver operating characteristic analysis. Radiology 1997;205:859–63.[Abstract/Free Full Text]
- Puhakka T, Mäkelä MJ, Alanen A, Kallio T, Korsoff L, Arstila P, et al. Sinusitis in the common cold. J Allergy Clin Immunol 1998;102:403–8.
- Sterling L, Tait GA, Edmonds JF. Interpretation of digital radiographs by pediatric critical care physicians using Web-based bedside personal computers versus diagnostic workstations. Pediatr Crit Care Med 2003;4:26–32.[CrossRef][Medline]
- Cox GG, Cook LT, McMillan JH, Rosenthal SJ, Dwyer SJ 3rd. Chest radiography: comparison of high-resolution digital displays with conventional and digital film. Radiology 1990;176:771–6.[Abstract/Free Full Text]
Top
Abstract
Introduction
