Learning to segment with limited annotations: Self-supervised pretraining with regression and contrastive loss in MRI

TL;DR

This paper explores self-supervised pretraining methods, specifically regression and contrastive loss, to improve MRI segmentation performance with limited labeled data, showing contrastive pretraining yields better results.

Contribution

It introduces and evaluates two self-supervised pretraining approaches for MRI segmentation, highlighting the effectiveness of contrastive loss over regression loss.

Findings

Self-supervised pretraining enables effective MRI segmentation with fewer labels.

Contrastive loss pretraining outperforms regression loss in downstream tasks.

Pretrained models achieve comparable performance to fully supervised models with limited annotations.

Abstract

Obtaining manual annotations for large datasets for supervised training of deep learning (DL) models is challenging. The availability of large unlabeled datasets compared to labeled ones motivate the use of self-supervised pretraining to initialize DL models for subsequent segmentation tasks. In this work, we consider two pre-training approaches for driving a DL model to learn different representations using: a) regression loss that exploits spatial dependencies within an image and b) contrastive loss that exploits semantic similarity between pairs of images. The effect of pretraining techniques is evaluated in two downstream segmentation applications using Magnetic Resonance (MR) images: a) liver segmentation in abdominal T2-weighted MR images and b) prostate segmentation in T2-weighted MR images of the prostate. We observed that DL models pretrained using self-supervision can be finetuned for comparable performance with fewer labeled datasets. Additionally, we also observed that initializing the DL model using contrastive loss based pretraining performed better than the regression loss.