Abstract

Traditional computer-aided detection (CAD) systems for breast cancer screening relied on machine learning with human-coded feature-engineering. They have largely failed to fulfil the promise of improving screening accuracy and workflow efficiency, and are often associated with increased recall rates and avoidable screening costs due to high instances of false positive markings. Advances in machine learning (such as deep learning) are on the cusp of providing more effective, more efficient, and even more patient-centric breast cancer screening support than ever before. By leveraging the consistent high sensitivity and specificity performance of autonomous systems, in combination with expert human oversight, the potential for efficient single-reader software-supported screening programs with low recall rates is on the horizon.