![[2602.20289] The Sim-to-Real Gap in MRS Quantification: A Systematic Deep Learning Validation for GABA](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.20289] The Sim-to-Real Gap in MRS Quantification: A Systematic Deep Learning Validation for GABA
Summary
This article presents a systematic validation of deep learning techniques for quantifying GABA in magnetic resonance spectroscopy (MRS), addressing challenges related to low signal-to-noise ratios and spectral overlap.
Why It Matters
The research highlights advancements in using deep learning to improve the quantification of metabolites in MRS, which is crucial for diagnosing neurological disorders and cancers. By bridging the sim-to-real gap, this study enhances the reliability of MRS data interpretation, potentially impacting clinical practices.
Key Takeaways
- Deep learning models significantly improve GABA quantification in MRS compared to traditional methods.
- Physics-informed data augmentation reduces the sim-to-real gap in model performance.
- The study emphasizes the importance of phantom ground truth for reliable MRS data interpretation.
Electrical Engineering and Systems Science > Signal Processing arXiv:2602.20289 (eess) [Submitted on 23 Feb 2026] Title:The Sim-to-Real Gap in MRS Quantification: A Systematic Deep Learning Validation for GABA Authors:Zien Ma, S. M. Shermer, Oktay Karakuş, Frank C. Langbein View a PDF of the paper titled The Sim-to-Real Gap in MRS Quantification: A Systematic Deep Learning Validation for GABA, by Zien Ma and 3 other authors View PDF HTML (experimental) Abstract:Magnetic resonance spectroscopy (MRS) is used to quantify metabolites in vivo and estimate biomarkers for conditions ranging from neurological disorders to cancers. Quantifying low-concentration metabolites such as GABA ($\gamma$-aminobutyric acid) is challenging due to low signal-to-noise ratio (SNR) and spectral overlap. We investigate and validate deep learning for quantifying complex, low-SNR, overlapping signals from MEGA-PRESS spectra, devise a convolutional neural network (CNN) and a Y-shaped autoencoder (YAE), and select the best models via Bayesian optimisation on 10,000 simulated spectra from slice-profile-aware MEGA-PRESS simulations. The selected models are trained on 100,000 simulated spectra. We validate their performance on 144 spectra from 112 experimental phantoms containing five metabolites of interest (GABA, Glu, Gln, NAA, Cr) with known ground truth concentrations across solution and gel series acquired at 3 T under varied bandwidths and implementations. These models are further assessed against ...
