Evaluating and Calibrating Diffusion Model-derived Uncertainty for Quantitative MRI Mapping

2026-08-12

AI Summary

This work addresses the lack of interpretable and quantifiable uncertainty estimation in existing deep learning–based quantitative MRI (qMRI) methods. Building upon a data-consistent diffusion model framework, the study generates uncertainty maps through multiple inference passes and, for the first time, systematically evaluates their correlation with ground-truth errors. A novel posterior calibration strategy is introduced, integrating bias correction with uncertainty scaling to substantially enhance the quantitative interpretability of prediction intervals. Experiments demonstrate a strong correlation between estimated uncertainty and actual mapping errors, showing that excluding high-uncertainty voxels significantly reduces error in the retained regions. Furthermore, the calibrated uncertainty intervals exhibit spatially plausible and statistically reliable coverage properties on healthy volunteer data.

Abstract

Quantitative MRI (qMRI) provides standardised tissue parameter maps, but the reliability of deep learning-based qMRI mapping methods is often not explicitly characterised. In this work we systematically evaluate uncertainty maps for quantitative MRI derived from multiple inferences of a data-consistent diffusion model-based qMRI framework. Evaluation on synthetic test data assessed error-awareness, high-error detection, selective prediction, and Gaussian interval calibration. Diffusion model-derived uncertainty was positively associated with the mapping error, while risk-coverage analysis showed that excluding high-uncertainty voxels reduced the retained error. However, the raw uncertainty was poorly calibrated for quantitative interval interpretation. Calibration was substantially improved using a post-hoc procedure combining prediction-value-dependent bias correction with scalar uncertainty scaling. Qualitative evaluation on a healthy volunteer showed spatially meaningful uncertainty patterns. These results indicate that diffusion model-derived uncertainty is informative for reliability assessment and selective prediction, but requires calibration for quantitative interval interpretation.

Problem

• quantitative MRI
• uncertainty estimation
• diffusion models
• calibration
• reliability assessment

Innovation

• diffusion model
• uncertainty calibration
• quantitative MRI
• selective prediction
• error-awareness