How Should We Evaluate Uncertainty in Accelerated MRI Reconstruction?

2025-03-13

AI Summary

Current uncertainty quantification in accelerated MRI reconstruction relies predominantly on pixel-level metrics (e.g., PSNR, SSIM), neglecting anatomical consistency and clinical task relevance. Method: We propose a novel anatomy-centric uncertainty quantification paradigm, introducing non-rigid image registration and medical image segmentation into reconstruction evaluation for the first time. Our framework—termed “anatomical consistency assessment”—integrates deep learning-based reconstruction, model ensembling, registration, and segmentation to quantify structural variations, including organ deformation and boundary displacement. Contribution/Results: Experiments reveal that state-of-the-art models achieving high PSNR/SSIM exhibit significant, systematic anatomical biases—compromising downstream segmentation accuracy and diagnostic reliability. This work establishes a task-oriented, interpretable, and anatomy-aware standard for uncertainty evaluation, advancing clinical trustworthiness of AI in medical imaging.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Uncertainty Representations
• Computer Vision: Segmentation

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Reconstructing accelerated MRI is an ill-posed problem. Machine learning has recently shown great promise at this task, but current approaches to quantifying uncertainty focus on measuring the variability in pixelwise intensity variation. Although these provide interpretable maps, they lack structural understanding and they do not have a clear relationship to how the data will be analysed subsequently. In this paper, we propose a new approach to evaluating reconstruction variability based on apparent anatomical changes in the reconstruction, which is more tightly related to common downstream tasks. We use image registration and segmentation to evaluate several common MRI reconstruction approaches, where uncertainty is measured via ensembling, for accelerated imaging. We demonstrate the intrinsic variability in reconstructed images and show that models with high scores on often used quality metrics such as SSIM and PSNR, can nonetheless display high levels of variance and bias in anatomical measures.

Problem

• Evaluating uncertainty in accelerated MRI reconstruction
• Linking uncertainty to anatomical changes and downstream tasks
• Assessing variability and bias in reconstructed anatomical measures

Innovation

• Evaluates MRI reconstruction via anatomical changes
• Uses image registration and segmentation techniques
• Measures uncertainty through ensembling in MRI