Optimal Latent Vector Alignment for Unsupervised Domain Adaptation in Medical Image Segmentation

TL;DR

This paper introduces OLVA, a lightweight unsupervised domain adaptation method using VAE and Optimal Transport to improve medical image segmentation across different imaging modalities.

Contribution

OLVA is a novel approach that learns a shared latent space focusing on shape, aligning domains with OT, and significantly enhances segmentation performance in multi-modality medical images.

Findings

Achieves 12.5% higher dice score over existing methods

Effectively aligns domain distributions in latent space

Improves segmentation accuracy on cardiac structures

Abstract

This paper addresses the domain shift problem for segmentation. As a solution, we propose OLVA, a novel and lightweight unsupervised domain adaptation method based on a Variational Auto-Encoder (VAE) and Optimal Transport (OT) theory. Thanks to the VAE, our model learns a shared cross-domain latent space that follows a normal distribution, which reduces the domain shift. To guarantee valid segmentations, our shared latent space is designed to model the shape rather than the intensity variations. We further rely on an OT loss to match and align the remaining discrepancy between the two domains in the latent space. We demonstrate OLVA's effectiveness for the segmentation of multiple cardiac structures on the public Multi-Modality Whole Heart Segmentation (MM-WHS) dataset, where the source domain consists of annotated 3D MR images and the unlabelled target domain of 3D CTs. Our results show remarkable improvements with an additional margin of 12.5\% dice score over concurrent generative training approaches.