Evaluation of interstitial lung diseases by use of temporal subtraction technique on computed radiography (CR) chest images. Detection and recognition of simulated lesions

Temporal subtraction images are obtained by the subtraction of a previous image from a current image. We investigated the detection of simulated lesions and also performed a recognition task in interstitial lung diseases using CR (computed radiography) images and temporal subtraction images. Five types of lung lesions, namely, ground-glass, reticular (no.1 and no.2), honeycomb, and micro-nodule patterns were simulated. Each simulated lesion was superimposed on one of the left lung, right lung, or mediastinum. Chest phantom images without and with simulated lesions were radiographed as previous and current images, respectively. Seventy-five CR and temporal subtraction images for each independent condition were used for evaluation. Five radiologists subjectively evaluated the detection and recognition of simulated lesions on CR images and temporal subtraction images. The results showed that the detection and recognition of simulated interstitial lung lesions on temporal subtraction images was significantly improved compared with CR images. Furthermore, the high detection rate was obtained with temporal subtraction images regardless of the subtlety and location of simulated lesions. (author)