Uncertainty aware training to improve deep learning model calibration for classification of cardiac MR images

Tareen Dawood, Chen Chen, Baldeep S. Sidhu, Bram Ruijsink, Justin Gould, Bradley Porter, Mark K. Elliott, Vishal Mehta, Christopher A. Rinaldi, Esther Puyol-Antón, Reza Razavi, Andrew P. King

Abstract: Highlights•Propose three novel uncertainty-aware training strategies.•Compare with state-of-the-art methods on two medical image classification tasks.•Report a wide range of accuracy and calibration performance measures.•Quantify performance of all models in terms of aleatoric and epistemic uncertainty.•Evaluate use of calibration-based model selection on accuracy and calibration.