Open World MRI Reconstruction with Bias-Calibrated Adaptation

2026-03-13

AI Summary

This work addresses the significant performance degradation of MRI reconstruction methods in real-world scenarios involving unknown imaging centers, anatomical structures, or acquisition protocols. To tackle this challenge, the authors propose BiasRecon, a test-time adaptation framework grounded in the principle of minimal intervention, which calibrates non-transferable biases while preserving transferable features for open-world generalization. The method introduces a novel frequency-guided prior calibration mechanism and an adaptive regularization strategy, requiring fewer than 100 tunable parameters. By integrating inter-layer calibration variables, score-based denoising, and a Stein unbiased risk estimator, BiasRecon enables efficient self-supervised optimization using k-space signals alone. Evaluated across four datasets, the approach achieves state-of-the-art performance in open-world MRI reconstruction and substantially enhances cross-domain generalization.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Bias, Fairness & Privacy
• Search and Optimization: Non-convex Optimization

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Real-world MRI reconstruction systems face the open-world challenge: test data from unseen imaging centers, anatomical structures, or acquisition protocols can differ drastically from training data, causing severe performance degradation. Existing methods struggle with this challenge. To address this, we propose BiasRecon, a bias-calibrated adaptation framework grounded in the minimal intervention principle: preserve what transfers, calibrate what does not. Concretely, BiasRecon formulates open-world adaptation as an alternating optimization framework that jointly optimizes three components: (1) frequency-guided prior calibration that introduces layer-wise calibration variables to selectively modulate frequency-specific features of the pre-trained score network via self-supervised k-space signals, (2) score-based denoising that leverages the calibrated generative prior for high-fidelity image reconstruction, and (3) adaptive regularization that employs Stein's Unbiased Risk Estimator to dynamically balance the prior-measurement trade-off, matching test-time noise characteristics without requiring ground truth. By intervening minimally and precisely through this alternating scheme, BiasRecon achieves robust adaptation with fewer than 100 tunable parameters. Extensive experiments across four datasets demonstrate state-of-the-art performance on open-world reconstruction tasks.

Problem

• Open-world MRI reconstruction
• domain shift
• bias calibration
• generalization
• test-time adaptation

Innovation

• open-world MRI reconstruction
• bias-calibrated adaptation
• score-based generative model
• frequency-guided calibration
• Stein's Unbiased Risk Estimator