RemInD: Remembering Anatomical Variations for Interpretable Domain Adaptive Medical Image Segmentation

2025-02-15

AI Summary

To address weak structural interpretability and poor generalizability of high-dimensional features in unsupervised domain adaptation (UDA) for medical imaging, this paper introduces a human-inspired memory mechanism. It models target-domain variations on a low-dimensional anatomical anchor manifold and proposes a dual-factor prediction framework constrained by probabilistic simplices—jointly learning interpretable anchor-weight vectors and spatial deformation fields. The method integrates Bayesian modeling, anchor-driven manifold learning, and deformation-field optimization to ensure geometric consistency and computational efficiency. Evaluated on multi-center public datasets for cardiac and abdominal imaging, it achieves state-of-the-art performance with only a single-stage alignment—outperforming existing methods reliant on complex, multi-stage strategies. The approach significantly enhances clinical interpretability and deployment feasibility while preserving anatomical fidelity.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Medical and Biological Imaging
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

This work presents a novel Bayesian framework for unsupervised domain adaptation (UDA) in medical image segmentation. While prior works have explored this clinically significant task using various strategies of domain alignment, they often lack an explicit and explainable mechanism to ensure that target image features capture meaningful structural information. Besides, these methods are prone to the curse of dimensionality, inevitably leading to challenges in interpretability and computational efficiency. To address these limitations, we propose RemInD, a framework inspired by human adaptation. RemInD learns a domain-agnostic latent manifold, characterized by several anchors, to memorize anatomical variations. By mapping images onto this manifold as weighted anchor averages, our approach ensures realistic and reliable predictions. This design mirrors how humans develop representative components to understand images and then retrieve component combinations from memory to guide segmentation. Notably, model prediction is determined by two explainable factors: a low-dimensional anchor weight vector, and a spatial deformation. This design facilitates computationally efficient and geometry-adherent adaptation by aligning weight vectors between domains on a probability simplex. Experiments on two public datasets, encompassing cardiac and abdominal imaging, demonstrate the superiority of RemInD, which achieves state-of-the-art performance using a single alignment approach, outperforming existing methods that often rely on multiple complex alignment strategies.

Problem

• Unsupervised domain adaptation in medical imaging
• Interpretable anatomical variations memorization
• Efficient and reliable image segmentation

Innovation

• Bayesian framework UDA
• Domain-agnostic latent manifold
• Explainable anchor weight vector