Interaction of a priori Anatomic Knowledge with Self-Supervised Contrastive Learning in Cardiac Magnetic Resonance Imaging
This work addresses the problem of improving diagnostic accuracy in cardiac MRI for medical imaging researchers, but it is incremental as it builds on existing SSCL methods without introducing a major breakthrough.
The study tackled the challenge of training deep learning models on cardiac MRI with limited labels by evaluating how to incorporate prior anatomic knowledge into self-supervised contrastive learning, finding that using a segmentation network to localize the heart improved downstream diagnostic performance, with SSCL pre-training on in-domain data outperforming end-to-end training and ImageNet pre-trained networks.
Training deep learning models on cardiac magnetic resonance imaging (CMR) can be a challenge due to the small amount of expert generated labels and inherent complexity of data source. Self-supervised contrastive learning (SSCL) has recently been shown to boost performance in several medical imaging tasks. However, it is unclear how much the pre-trained representation reflects the primary organ of interest compared to spurious surrounding tissue. In this work, we evaluate the optimal method of incorporating prior knowledge of anatomy into a SSCL training paradigm. Specifically, we evaluate using a segmentation network to explicitly local the heart in CMR images, followed by SSCL pretraining in multiple diagnostic tasks. We find that using a priori knowledge of anatomy can greatly improve the downstream diagnostic performance. Furthermore, SSCL pre-training with in-domain data generally improved downstream performance and more human-like saliency compared to end-to-end training and ImageNet pre-trained networks. However, introducing anatomic knowledge to pre-training generally does not have significant impact.