Transfusion: Understanding Transfer Learning for Medical Imaging

TL;DR

This paper investigates transfer learning in medical imaging, revealing that it often offers limited benefits and that simpler models can perform similarly to complex pretrained architectures, emphasizing the importance of model efficiency.

Contribution

The study provides a detailed analysis of transfer learning effects in medical imaging, highlighting when and where transfer is beneficial and suggesting more efficient model exploration strategies.

Findings

Transfer learning offers limited performance gains in medical imaging.

Simple, lightweight models can match complex pretrained models.

Over-parametrization, not feature reuse, explains differences in transfer learning effectiveness.

Abstract

Transfer learning from natural image datasets, particularly ImageNet, using standard large models and corresponding pretrained weights has become a de-facto method for deep learning applications to medical imaging. However, there are fundamental differences in data sizes, features and task specifications between natural image classification and the target medical tasks, and there is little understanding of the effects of transfer. In this paper, we explore properties of transfer learning for medical imaging. A performance evaluation on two large scale medical imaging tasks shows that surprisingly, transfer offers little benefit to performance, and simple, lightweight models can perform comparably to ImageNet architectures. Investigating the learned representations and features, we find that some of the differences from transfer learning are due to the over-parametrization of standard models rather than sophisticated feature reuse. We isolate where useful feature reuse occurs, and outline the implications for more efficient model exploration. We also explore feature independent benefits of transfer arising from weight scalings.