Abstract
A common approach for probing brain microstructure is the acquisition of multi-contrast quantitative MRI, including diffusion MRI (dMRI) and multi-parameter mapping (MPM) of relaxation rates and proton density (PD). Quantitative $\mathrm{T}_2$ mapping, however, is often omitted from in-vivo protocols due to time constraints. Thus, clinically relevant maps such as $\mathrm{R}_2^\prime$ cannot be be calculated although these could for example provide insights in iron deposition as well as myelin via magnetic susceptibility source separation. We propose a neural network based approach for estimating the $\mathrm{R}_2$ map from PD, $\mathrm{R}_1$ and $\mathrm{R}_2^*$ maps as well as $\mathrm{B}_0$ fieldmaps typically obtained alongside MPM measurements. In addition, we investigated whether providing the network a starting estimate modeled from diffusion $b = 0$ signal, PD, and $\mathrm{R}_1$ improves prediction performance. The intended use case is incomplete datasets in which MPM and diffusion MRI have already been acquired but dedicated quantitative $\mathrm{T}_2$ mapping is unavailable. We compared the network predictions against reference $\mathrm{R}_2$ maps from dedicated $\mathrm{T}_2$ measurements. Across subject-wise cross-validation, providing the model-based starting estimate reduced systematic bias in white and gray matter by approximately $60$ % and $68$ %, respectively, compared with a MPM-only U-Net, while reductions in overall error were smaller. These results indicate that an already available diffusion $b = 0$ image can improve the estimation of missing $\mathrm{R}_2$ maps. However, residual errors remain substantial, and larger independent datasets are required before predicted $\mathrm{R}_2$ maps can be considered quantitative biomarkers.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/cdmri_016.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=ig0p2pftWh
BibTex
@InProceedings{MeyJan_Quantitative_MICCAISAT2026,
author = { Meyer, Jan AND Bogs, Laura AND Lüthi, Nina AND Fritz, Francisco J. AND Sura, Noémie AND Oeschger, Jan Malte AND Mordhorst, Laurin AND Sauvigny, Thomas AND Heinrich, Mattias P. AND Natho-Mohammadi, Siawoosh},
title = { { Quantitative R2 Estimation from Multi-Parameter and dMRI Measurements using Neural Networks } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17278},
month = {pending},
page = {pending}
}