ELECTRIC: Evidential Learning-Enhanced CT Reconstruction via Iterative Correction

2026-07-27

AI Summary

This work addresses reconstruction errors and data inconsistency in low-dose CT imaging caused by inaccurate priors by proposing a physics-guided Bayesian iterative reconstruction framework. The method uniquely incorporates epistemic uncertainty, derived from evidential deep learning, as a learnable state variable to construct an adaptive precision field, which is integrated into a Poisson-weighted maximum a posteriori (MAP) update. This forms a closed-loop mechanism—image–evidence–precision—that jointly and adaptively optimizes prior strength and data fidelity. Evaluated on the AAPM Mayo dataset, the approach reduces reconstruction error by approximately 70% compared to filtered back-projection. Moreover, the estimated epistemic uncertainty effectively predicts reconstruction error and demonstrates robustness under prior misspecification, significantly outperforming methods employing fixed priors.

Abstract

Here we introduce ELECTRIC (Evidential Learning-Enhanced CT Reconstruction via Iterative Correction), a physics-guided Bayesian formulation. An evidential neural network provides an image proposal and an error-predictive epistemic-uncertainty surrogate. The latter is converted into an adaptive precision field and inserted into a Poisson-weighted MAP update. The resulting image-evidence-precision-reconstruction loop treats prior confidence as a learned state variable of iterative reconstruction. In addition to the formulation and theoretical analysis, we report two simulation studies on image slices from the AAPM Mayo Clinic Low-Dose CT dataset: a mechanism-validation pilot using transparent surrogate estimators, and a feasibility study in which a trained Normal-Inverse-Gamma evidential network drives the full closed loop. On held-out patients, the learned prior mean reduces reconstruction error by roughly 70 percent relative to filtered back-projection, the learned epistemic uncertainty is error-predictive and supports selective trust, and the physics-guided update restores measurement consistency while the adaptive-precision reconstruction matches or exceeds a validation-tuned fixed prior and remains markedly more robust to prior-strength misspecification. Together these results demonstrate the complete ELECTRIC closed-loop pipeline, while identifying formal uncertainty calibration and joint training as the principal directions for future work.

Problem

• CT reconstruction
• epistemic uncertainty
• low-dose CT
• Bayesian inference
• physics-guided learning

Innovation

• evidential learning
• adaptive precision
• Bayesian CT reconstruction
• epistemic uncertainty
• physics-guided iterative correction