Abstract
Magnetic Resonance Spectroscopy (MRS) quantification faces persistent, long-standing significant challenges due to the inherent limitations of conventional parametric methods, as they rely on strict and rigid assumptions related to metabolite basis signals, spectral line-shapes, macromolecule profiles, baselines, and other factors. Recently, deep learning methods have been explored as powerful and innovative alternatives, showing promising results. However, the considerable and intrinsic variability in physical parameters, such as amplitudes, frequency shifts, damping factors, macromolecule signals, and baselines, remains a critical obstacle. For the practicality of the deployment in clinical routine or to reduce the acquisition time using advanced scheme, it is important for the estimation to be not only robust and accurate but also lightweight and fast. To address this challenge, we introduce a novel deep learning framework designed to capture a wide spectrum of fluctuations in MRS signals using two steps in the estimation. A coarse prediction is used to help the final estimation of the parameters in the form of a network conditioning. Experimental results demonstrate that although our framework is lighter, it achieves similar performance and sometimes outperforms larger networks in accuracy and robustness of metabolite quantification, even in the presence of significant parameter variability.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MLMI_077.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: https://papers.miccai.org/miccai-2026-sat/supp/MLMI_077_supp.pdf
Link to Open Review
Open Review Page: https://openreview.net/forum?id=xvUPBfHJNo
BibTex
@InProceedings{HacSli_Acoarse_MICCAISAT2026,
author = { Hachicha, Slim AND Ratiney, Helene AND Sdika, Michaël},
title = { { A coarse to fine approach to magnetic resonance spectroscopy quantification } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17272},
month = {pending},
page = {pending}
}