Quadtree-based data compression of abdominal CT images

An ROC study was performed to evaluate the effect of quadtree-based data compression on the diagnostic yield of CT images. Seventy images were selected from a series of consecutive abdominal/pelvic CT scans. Following the application of quadtree-based compression, all images were reviewed independently by three radiologists. The images were analyzed at six decreasing irreversible compression ratios (30.6:1 to 7.4:1), as well as after reversible compression (2.9:1). ROC curves reveal a gradual decrease in clinical accuracy with increasing compression ratios. At a compression ratio of 7.4:1, sensitivity for all major abnormalities was 99% with a specificity of 93%. As the compression ratio was increased to 30.6:1, sensitivity and specificity dropped to 75% and 83% respectively. Execution times for compression and decompression of the CT images with a PC-AT based digital radiography system varied from 24.7 to 18.5 seconds and from 16.2 to 5.1 seconds respectively, decreasing with higher levels of compression. We conclude that quadtree-based compression of abdominal CT images may have practical applications for a PC based digital radiography system.