Deep learning for MRI and ultrasound

Deep learning work applied to MRI and ultrasound, carried out at Polytechnique Montréal.

One of these projects, presented at ISMRM 2021, explores deep learning super-resolution for quantitative susceptibility mapping (QSM), on the AHEAD dataset. It compares several mean-squared-error loss variants and training strategies: a cyclic learning rate generally yields better results when correctly tuned with the learning rate finder algorithm.