Complex Style Image Transformations for Domain Generalization in Medical Images

TL;DR

This paper introduces CompStyle, a novel framework combining style transfer, adversarial training, and input augmentation to improve domain generalization in medical image analysis, especially when data is scarce or diverse.

Contribution

The paper presents a new method that effectively expands domain space and enhances model robustness without increasing training costs.

Findings

Improved semantic segmentation accuracy on prostate data.

Enhanced robustness to image corruptions on cardiac data.

No additional training time or resource requirements.

Abstract

The absence of well-structured large datasets in medical computer vision results in decreased performance of automated systems and, especially, of deep learning models. Domain generalization techniques aim to approach unknown domains from a single data source. In this paper we introduce a novel framework, named CompStyle, which leverages style transfer and adversarial training, along with high-level input complexity augmentation to effectively expand the domain space and address unknown distributions. State-of-the-art style transfer methods depend on the existence of subdomains within the source dataset. However, this can lead to an inherent dataset bias in the image creation. Input-level augmentation can provide a solution to this problem by widening the domain space in the source dataset and boost performance on out-of-domain distributions. We provide results from experiments on semantic segmentation on prostate data and corruption robustness on cardiac data which demonstrate the effectiveness of our approach. Our method increases performance in both tasks, without added cost to training time or resources.