Logo image
Dual Mixed Style Augmentation for Generalized Medical Image Segmentation
Conference proceeding

Dual Mixed Style Augmentation for Generalized Medical Image Segmentation

Jiaxi Li, Yuntong Tian, Liang Wan and ASSOC COMPUTING MACHINERY
Proceedings of the 2023 6th International Conference on Artificial Intelligence and Pattern Recognition, pp.134-139
ACM Other Conferences
AIPR 2023: 2023 6th International Conference on Artificial Intelligence and Pattern Recognition
22/09/2023

Abstract

Applied computing Applied computing -- Life and medical sciences Applied computing -- Life and medical sciences -- Computational biology Applied computing -- Life and medical sciences -- Computational biology -- Imaging Applied computing -- Life and medical sciences -- Health care information systems Computing methodologies Computing methodologies -- Artificial intelligence Computing methodologies -- Artificial intelligence -- Computer vision Computing methodologies -- Artificial intelligence -- Computer vision -- Computer vision problems Computing methodologies -- Artificial intelligence -- Computer vision -- Computer vision problems -- Image segmentation Computing methodologies -- Artificial intelligence -- Computer vision -- Computer vision problems -- Video segmentation Computing methodologies -- Machine learning Computing methodologies -- Machine learning -- Learning paradigms Computing methodologies -- Machine learning -- Learning paradigms -- Unsupervised learning
Domain generalization aims to generalize models trained on source domains to unseen target domains, which usually have domain shift from the source domains. This challenging setting is common in clinical scenarios due to differences in scanning protocols, device manufacturers, and image modalities. In this paper, a novel style augmentation framework is proposed to improve out-of-domain performance for the model in the context of medical image segmentation. The proposed framework first investigates the style distribution between samples, for which a style clustering-based resampling is developed to generate evenly diversified styles, and then leverages a cross-structure style augmentation, which generates new styles by shuffling regional style information among structure within one sample. This structure style shuffling can greatly expand the style space while well respecting the anatomical structure of medical images. The method is validated through extensive experiments on 2 datasets, i.e., cross-modality abdominal multi-organ dataset, and cross-site cardiac dataset, with 2 settings, i.e., single-domain and multi-domain generalization. Experimental results show that our method yields significant improvement and outperforms state-of-the-art methods with significant margins.

Metrics

1 Record Views

Details

Logo image