A Comprehensive Framework for Uncertainty Quantification of Voxel‐Wise Supervised Deep Learning Models in IVIM MRI
NMR in Biomedicine · 2026
DOI: 10.1002/nbm.70227 ↗Kurz gefasst
Fachartikel in der Zeitschrift NMR in Biomedicine, 2026. Verfasst von Nicola Casali und 8 weiteren Autor:innen. Beteiligte Einrichtung: Istituto di Sistemi e Tecnologie Industriali Intelligenti per il Manifatturiero Avanzato, Consiglio Nazionale delle Ricerche, Milan, Italy. Hauptthemen laut MEDLINE: Deep Learning, Supervised Machine Learning, Magnetic Resonance Imaging, Diffusion Magnetic Resonance Imaging. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 59 Literaturangaben. Gefördert von: Italian Ministry of University and Research, Fondazione Regionale per la Ricerca Biomedica (Regione Lombardia), Regione Lombardia.
Abstract (Originalsprache)
Accurate estimation of intravoxel incoherent motion (IVIM) parameters from diffusion‐weighted MRI remains challenging due to the ill‐posed nature of the inverse problem and high sensitivity to noise, particularly in the perfusion compartment. In this work, we propose a probabilistic deep learning framework based on deep ensembles (DEs) of mixture density networks (MDNs), enabling estimation of total predictive uncertainty and decomposition into aleatoric (AU) and epistemic (EU) components. The method was benchmarked against nonprobabilistic neural networks, a Bayesian fitting approach, and a probabilistic network with single Gaussian parametrization. Supervised training was performed on synthetic data, and evaluation was conducted on both simulated and in vivo brain mouse dataset. The reliability of the quantified uncertainties was assessed using calibration curves, output distribution sharpness, and the continuous ranked probability score (CRPS). MDNs produced more calibrated and sharper predictive distributions for the diffusion coefficient () and the perfusion fraction () parameters, although slight overconfidence was observed in the pseudodiffusion coefficient (). The robust coefficient of variation (RCV) indicated smoother in vivo estimates for with MDNs compared with Gaussian model. Despite the training data covering the expected physiological range, elevated EU in vivo suggests a mismatch with real acquisition conditions, highlighting the importance of incorporating EU, which was allowed by DE. Overall, we present a comprehensive framework for IVIM fitting with uncertainty quantification, which enables the identification and interpretation of unreliable estimates. The proposed approach can also be adopted for fitting other physical models through appropriate architectural and simulation adjustments.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Istituto di Sistemi e Tecnologie Industriali Intelligenti per il Manifatturiero Avanzato, Consiglio Nazionale delle Ricerche, Milan, Italy.
- Politecnico di Milano, Milan, Italy.
- Department of Acute Brain and Cardiovascular Injury, Istituto di Ricerche Farmacologiche Mario Negri IRCCS, Milan, Italy.
- Department of Neuroscience, Istituto di Ricerche Farmacologiche Mario Negri IRCCS, Milan, Italy.
- Division of Pulmonary Medicine, Cincinnati Children's Hospital Medical Center, Cincinnati, Ohio, USA.
ORCID-Profile: Nicola Casali, Alessandro Brusaferri, Giuseppe Baselli, Stefano Fumagalli, Edoardo Micotti, Gianluigi Forloni, Riaz Hussain, Giovanna Rizzo, Alfonso Mastropietro
Förderung
- Italian Ministry of University and Research
- Fondazione Regionale per la Ricerca Biomedica (Regione Lombardia)
- Regione Lombardia
Zitieren
APA
Casali, N., Brusaferri, A., Baselli, G., Fumagalli, S., Micotti, E., Forloni, G., Hussain, R., Rizzo, G., & Mastropietro, A. (2026). A Comprehensive Framework for Uncertainty Quantification of Voxel‐Wise Supervised Deep Learning Models in IVIM MRI. NMR in Biomedicine. https://doi.org/10.1002/nbm.70227
Vancouver
Casali N, Brusaferri A, Baselli G, Fumagalli S, Micotti E, Forloni G, et al. A Comprehensive Framework for Uncertainty Quantification of Voxel‐Wise Supervised Deep Learning Models in IVIM MRI. NMR in Biomedicine. 2026. doi:10.1002/nbm.70227
BibTeX
@article{Casali2026,
author = {Casali, Nicola and Brusaferri, Alessandro and Baselli, Giuseppe and Fumagalli, Stefano and Micotti, Edoardo and Forloni, Gianluigi and Hussain, Riaz and Rizzo, Giovanna and Mastropietro, Alfonso},
title = {A Comprehensive Framework for Uncertainty Quantification of Voxel‐Wise Supervised Deep Learning Models in IVIM MRI},
journal = {NMR in Biomedicine},
year = {2026},
doi = {10.1002/nbm.70227}
}
EUVIMED