Reliable Source Approximation: Source-Free Unsupervised Domain Adaptation for Vestibular Schwannoma MRI Segmentation

TL;DR

This work tackles domain shift in vestibular schwannoma MRI segmentation across sequences using source-free unsupervised domain adaptation (SFUDA). It introduces Reliable Source Approximation (RSA), which combines edge-guided diffusion to generate source-like, structure-preserved target images, an uncertainty-aware segmentation model based on a Normal-Inverse-Gamma prior, and a prediction-consistency based selection to identify reliable generations for training. The approach yields substantial improvements over existing SFUDA methods on ceT1→hrT2 vestibular schwannoma segmentation, with centralized fine-tuning delivering the strongest gains. The proposed method advances practical cross-sequence medical image segmentation by enabling robust adaptation without target annotations, and code is publicly available.

Abstract

Source-Free Unsupervised Domain Adaptation (SFUDA) has recently become a focus in the medical image domain adaptation, as it only utilizes the source model and does not require annotated target data. However, current SFUDA approaches cannot tackle the complex segmentation task across different MRI sequences, such as the vestibular schwannoma segmentation. To address this problem, we proposed Reliable Source Approximation (RSA), which can generate source-like and structure-preserved images from the target domain for updating model parameters and adapting domain shifts. Specifically, RSA deploys a conditional diffusion model to generate multiple source-like images under the guidance of varying edges of one target image. An uncertainty estimation module is then introduced to predict and refine reliable pseudo labels of generated images, and the prediction consistency is developed to select the most reliable generations. Subsequently, all reliable generated images and their pseudo labels are utilized to update the model. Our RSA is validated on vestibular schwannoma segmentation across multi-modality MRI. The experimental results demonstrate that RSA consistently improves domain adaptation performance over other state-of-the-art SFUDA methods. Code is available at https://github.com/zenghy96/Reliable-Source-Approximation.

Figures (2)

• Figure 1: Overview of the proposed SFUDA method. Two models are pre-trained using source data, and the target adaptation phase consists of Reliable Source Approximation (RSA, black arrows) and model fine-tuning (green arrows).
• Figure 2: Visulazition of reliable source approximation and segmentation results.