The Effect of Irreversible Image Compression on Diagnostic Accuracy in Thoracic Imaging

Abstract

Digital image compression reduces the storage requirements and network traffic on picture archiving and communications systems. Full-frame bit-allocation (FFBA) is an irreversible image-compression method based on the discrete cosine transform that provides for high compression ratios with a high degree of image fidelity. One hundred twenty-two posteroanterior chest radiographs were obtained on patients in an ambulatory patient setting, including 30 cases of interstitial lung disease, 45 images containing combinations of lung nodules (N = 37) or mediastinal masses (N = 39), and 47 normal images containing none of the pathology for which we were testing. The images were digitized (nominal 2 K x 2 K x 12-bit resolution), printed on a 35 x 35-cm hard copy format, and compressed at an approximate compression ratio of 20:1. Observer performance tests were conducted with five radiologists using receiver operating characteristics analysis on digitized uncompressed and compressed hard copy images. There were no significant differences between the two display conditions for the detectability of any of the thoracic abnormalities. Our preliminary results suggest that irreversible image compression at ratios of 20:1 may be acceptable for use in digital thoracic imaging.