Mass detection on automated breast ultrasound volume scans using convolutional neural network

Automated ultrasound volume scan is useful for screening breast cancer as it records the whole breast data allowing longitudinal comparisons and double reading. Unlike examinations with handheld probes, it is less operator independent. However, the number of acquired images is large, which increases radiologists' workload. In this study, we propose a mass detection method on ultrasound volume scans using convolutional neural network (CNN). Using CNN, detection performance greatly improved compared with the conventional filter based method.