Fundamental limitations in developing computer-aided detection for mammography

Abstract

While asymptomatic screening with mammography has been proven to reduce breast cancer mortality, radiologists miss cancers when reading screening mammograms. Computer-aided detection (CADe) is being developed to help radiologists avoid overlooking a cancer. In this paper, we describe two overarching issues that limit the current development of CADe schemes. These are the inability to optimize a scheme for clinical impact – current methods only optimize for how well the CADe scheme works in the absence of a radiologist – and the lack of a figure of merit that quantifies the performance efficiency of the CADe scheme. Such a figure of merit could be used to determine how much better performance a CADe scheme could obtain, at least in theory, and which component of the several techniques employed in the CADe scheme is the weakest link.