CT colonography with teleradiology: effect of lossy wavelet compression on polyp detection--initial observations.

Abstract

To assess the consequences of lossy compression on the diagnostic accuracy of CT colonography for detecting colonic polyps. Helical CT images of cleansed colonic segments were evaluated. Source images were compressed to 1:1, 10:1, and 20:1 ratios with lossy wavelet compression. Two independent readers blinded to corresponding colonoscopic results analyzed 144 randomly ordered colonic segments in multiplanar and volume-rendered endoscopic views. Sensitivity, specificity, and receiver operating characteristic curves were generated for each compression ratio on the basis of expressed confidence in lesion presence. Similar analyses were performed to assess distention and bowel preparation adequacy and evaluation time required. Results based on video colonoscopy-confirmed lesions revealed 100% (four of four) sensitivity for lesions larger than 10 mm for compression ratios 1:1, 10:1, and 20:1 for both readers; sensitivities for all lesions smaller than 10 mm were 50%-78%, 38%-67%, and 38%-67% for respective ratios for both readers. Differences in diagnostic performance for each reader across ratios were not significant (P =.30-.99, McNemar test). The time required to evaluate and assess bowel preparation and distention adequacy did not change significantly across ratios. On the basis of the patient sample, lossy compression of transverse source images to at least a 20:1 ratio did not adversely affect diagnostic performance or evaluation time for CT colonography.