Aligning What You Separate: Denoised Patch Mixing for Source-Free Domain Adaptation in Medical Image Segmentation

TL;DR

This work tackles domain shift in medical image segmentation under privacy constraints by introducing a source-free domain adaptation strategy that progressively adapts from easy to hard target samples. It combines Hard Sample Selection with entropy and feature-similarity criteria, Monte Carlo-based pseudo-label denoising, and a Denoised Patch Mixing (DPM) mechanism to align distributions both within the reliable target subset and between reliable and unreliable regions. The training relies on a teacher-student framework with EMA updates and two loss channels (intra-domain and inter-domain BCE losses) to stabilize learning and suppress noisy supervision. Experiments on fundus image benchmarks demonstrate state-of-the-art Dice and ASSD scores, sharper boundaries, and robust performance under domain shift, highlighting the practical value of progressive, denoised supervision for privacy-preserving medical image segmentation.

Abstract

Source-Free Domain Adaptation (SFDA) is emerging as a compelling solution for medical image segmentation under privacy constraints, yet current approaches often ignore sample difficulty and struggle with noisy supervision under domain shift. We present a new SFDA framework that leverages Hard Sample Selection and Denoised Patch Mixing to progressively align target distributions. First, unlabeled images are partitioned into reliable and unreliable subsets through entropy-similarity analysis, allowing adaptation to start from easy samples and gradually incorporate harder ones. Next, pseudo-labels are refined via Monte Carlo-based denoising masks, which suppress unreliable pixels and stabilize training. Finally, intra- and inter-domain objectives mix patches between subsets, transferring reliable semantics while mitigating noise. Experiments on benchmark datasets show consistent gains over prior SFDA and UDA methods, delivering more accurate boundary delineation and achieving state-of-the-art Dice and ASSD scores. Our study highlights the importance of progressive adaptation and denoised supervision for robust segmentation under domain shift.

Figures (3)

• Figure 1: Performance comparison of different methods under the same backbone. The proposed framework significantly boosts Dice score compared to previous works.
• Figure 2: Overview of our proposed framework for source-free domain adaptive medical image segmentation.
• Figure 3: Qualitative comparisons of different methods.