A systematic review of computer-assisted diagnosis in diagnostic cancer imaging.

ABSTRACT

OBJECTIVES: This study reviews the evidence for the effectiveness of computer-assisted diagnosis (CAD) in cancer imaging. Diagnostic applications were studied to estimate the impact of CAD on radiologists' detection and diagnosis of cancer lesions. METHODS: Online databases were searched and 48 studies from 1992 to 2010 were included 16 with radiologists using CAD to detect lesions (CADe) and 32 with radiologists using CAD to classify or diagnose lesions (CADx). Weighted means, statistics, summary receiver operating characteristics (SROC) curves, and related measures were used for analysis. RESULTS: There is evidence that CADx significantly improves diagnosis in mammography and breast ultrasound. In contrast, studies of CADx applied to lung CT and dermatologic imaging show an adverse impact on diagnosis. Overall, there is no evidence of a benefit due to the use of CADe. The area under the SROC curves was not significantly increased for radiologists using either CADe or CADx. CONCLUSIONS: From this analysis it seems CADx can offer some benefit to radiologists in specific imaging applications for breast cancer diagnosis. There is no evidence of a beneficial effect in other applications of CAD and some evidence of a detrimental one.