Why does my medical AI look at pictures of birds? Exploring the efficacy of transfer learning across domain boundaries

TL;DR

Pretraining on domain-specific medical images like CT scans generally outperforms cross-domain pretraining on datasets like ImageNet, especially in transfer learning scenarios, due to domain boundary-related generalization gaps.

Contribution

This study systematically compares intra- and cross-domain transfer learning in medical imaging, highlighting the benefits of domain-specific pretraining and analyzing feature space differences.

Findings

Intra-domain transfer outperforms cross-domain transfer in most scenarios.

Pretraining on RadNet-12M yields up to 2.07% performance gains.

Domain boundaries influence generalization and feature representations.

Abstract

It is an open secret that ImageNet is treated as the panacea of pretraining. Particularly in medical machine learning, models not trained from scratch are often finetuned based on ImageNet-pretrained models. We posit that pretraining on data from the domain of the downstream task should almost always be preferred instead. We leverage RadNet-12M, a dataset containing more than 12 million computed tomography (CT) image slices, to explore the efficacy of self-supervised pretraining on medical and natural images. Our experiments cover intra- and cross-domain transfer scenarios, varying data scales, finetuning vs. linear evaluation, and feature space analysis. We observe that intra-domain transfer compares favorably to cross-domain transfer, achieving comparable or improved performance (0.44% - 2.07% performance increase using RadNet pretraining, depending on the experiment) and demonstrate the existence of a domain boundary-related generalization gap and domain-specific learned features.