Automated detection of mammographic masses: preliminary assessment of an information-theoretic CAD scheme for reduction of false positives

Abstract

ABSTRACT The purpose of this work was to evaluate an information-theoretic computer-aided detection (CAD) scheme forimproving the specificity of mass detection in screening mammograms. The study was based on images from theLumisys set of the Digital Database for Screening Mammography (DDSM). Initially, the craniocaudal views of 49DDSM mammograms were analyzed using an automated detection algorithm developed to prescreen mammograms.The prescreening algorithm followed a morphological concentric layer analysis and resulted in 319 false positivedetections at 92% sensitivity. These 319 suspicious yet normal regions were extracted for further analysis with ourinformation-theoretic CAD scheme. Our scheme follows a knowledge-based decision strategy. The strategy relies oninformation theoretic principles for similarity assessment between a query case and a knowledge databank of cases withknown ground truth. Receiver Operating Characteristic (ROC) analysis was performed to determine how well the CADscheme can discriminate the false positive regions from 681 true masses. The overall ROC area index of theinformation-theoretic CAD system was 0.75±0.02. At 97%, 95%, and 90% sensitivity, the system eliminated safely20%, 30%, and 42% of the previously identified false positives respectively. Thus, information-theoretic CAD analysiscan yield a significant reduction in false-positive detections while maintaining reasonable sensitivity.Keywords: computer-aided diagnosis, mammography, mass detection, mutual information.