EFFECTS OF JPEG AND WAVELET COMPRESSION OF SPIRAL LOW-DOSE CT IMAGES ON DETECTION OF SMALL LUNG CANCERS

Abstract

To compare the effect of compression of spiral low-dose CT images by the Joint Photographic Experts Group (JPEG) and wavelet algorithms on detection of small lung cancers. Low-dose spiral CT images of 104 individuals (52 with peripheral lung cancers smaller than 20 mm and 52 control subjects) were used. The original images were compressed using JPEG or wavelet algorithms at a ratio of 10:1 or 20:1. Five radiologists interpreted these images and evaluated the image quality on a high-resolution CRT monitor. Observer performance was studied by receiver operating characteristic (ROC) analysis. There was no significant difference in the detection of cancers measuring 6 to 15 mm in uncompressed images and in those compressed by either of the algorithms, although the quality of images compressed at 20:1 with the wavelet algorithm was somewhat inferior. A lower diagnostic accuracy was noted using images compressed by the JPEG or wavelet algorithms at 20:1 in detecting lung cancers measuring 6 to 10 mm and cancers measuring from 6 to 15 mm with ground-glass opacity. Compression of low-dose CT images at a ratio of 10:1 using JPEG and wavelet algorithms does not compromise the detection rate of small lung cancers.