Effect of Temporal Subtraction Technique on Interpretation Time and Diagnostic Accuracy of Chest Radiography

Abstract

Our purpose was to compare reviewing time and diagnostic accuracy in the interpretation of radiographs without and with subtraction images and to examine whether this temporal subtraction technique can contribute to improving radiologists' performance. Thirty cases with newly developed chest abnormalities on chest radiographs and 90 negative cases were selected. All chest radiographs were obtained with a computed radiography system. For the 90 negative cases, subtraction images were classified into two groups: 33 clean images without misregistration artifacts and 57 images with some misregistration artifacts. Eight radiologists (four board-certified radiologists and four radiology residents) participated in observer tests and interpreted the original radiographs without and with subtraction images using an independent test method. The reviewing time for each radiologist was recorded in each case. The observers' performance was evaluated by use of receiver operating characteristic (ROC) analysis. When subtraction images were available, the mean reviewing time per case was reduced significantly from 13.6 to 10.8 seconds for the cases with newly developed abnormalities (p < 0.001) and from 29.8 to 14.1 seconds for negative cases (p < 0.001). The reduction in the mean reviewing time with subtraction images was greater for clean images than for images with artifacts (17.7 vs 14.5 seconds, p < 0.001). The average mean area under the ROC curve value increased significantly from 0.942 without subtraction images to 0.988 with subtraction images (p = 0.025). There were significant differences in the sensitivity (0.963 with and 0.888 without the subtraction images) and the specificity (0.976 with and 0.899 without the subtraction images) (p < 0.001). The temporal subtraction technique can reduce reviewing time and also improve diagnostic accuracy in the interpretation of chest radiographs.