MixSearch: Searching for Domain Generalized Medical Image Segmentation Architectures

TL;DR

This paper introduces MixSearch, a NAS framework that searches for domain-generalized medical image segmentation architectures using a mixed dataset from multiple domains, achieving state-of-the-art results.

Contribution

MixSearch is the first NAS approach to focus on generalization across multiple medical image domains by creating a large composite dataset and designing a novel weaved encoder-decoder structure.

Findings

Achieves state-of-the-art segmentation performance across various datasets.

Effectively generalizes to unseen medical image domains.

Demonstrates the benefit of dataset mixing in NAS for medical imaging.

Abstract

Considering the scarcity of medical data, most datasets in medical image analysis are an order of magnitude smaller than those of natural images. However, most Network Architecture Search (NAS) approaches in medical images focused on specific datasets and did not take into account the generalization ability of the learned architectures on unseen datasets as well as different domains. In this paper, we address this point by proposing to search for generalizable U-shape architectures on a composited dataset that mixes medical images from multiple segmentation tasks and domains creatively, which is named MixSearch. Specifically, we propose a novel approach to mix multiple small-scale datasets from multiple domains and segmentation tasks to produce a large-scale dataset. Then, a novel weaved encoder-decoder structure is designed to search for a generalized segmentation network in both cell-level and network-level. The network produced by the proposed MixSearch framework achieves state-of-the-art results compared with advanced encoder-decoder networks across various datasets.