Computer-aided detection for mammography: improved algorithm performance with operator determined points characterized by new metrics

摘要: Abstract Recent prospective studies have suggested that user's confidence in computer-aided detection (CAD) marks and their tolerance of false positive (FP) play a role the benefit gained from CAD device. In this paper, we propose to (1) introduce new metrics improve characterization latest generation algorithms (2) describe algorithm performance tradeoffs with user adjustable operating points. The (V8.0, R2 Technology, Sunnyvale, CA) points was assessed on multi-institutional database 832 consecutive, biopsy-proven cancers detected by screening mammography 345 clinically confirmed normal (four view) cases. standard measurements were made—case sensitivity for microcalcifications masses, marker rate addition, case specificity (the percentage cases no marks) two-view cancer visible ( n =628) marked both views) used. It is likely measure reflects impact workflow radiologist. Further, high value increase radiologist's tool. This system allows choose an point best fit clinical requirements.