Active multitask learning with uncertainty‐weighted loss for coronary calcium scoring

Abstract

Our multitask learning approach with uncertainty-weighted loss improves calcium scoring performance by joint learning of shared features and reduces labeling costs when trained in an active learning scenario.