Accelerated Major Singlet Metabolite Mapping in Human Brain Using Nonwater/Fat‐Suppressed Ultrashort‐TR FID‐MRSI With FID‐Encoded IDEAL
Free-induction-decay MR spectroscopic imaging (FID-MRSI) with LCModel is a well-established, rapid brain metabolite mapping method. However, multislice whole brain coverage still entails substantial acquisition time. To enable water-referenced absolute mapping of the major brain singlets (total choline/creatine/N-acetylaspartate [tCho/tCr/tNAA]) with substantially reduced scan time, we developed nonwater/fat-suppressed ultrashort-TR FID-MRSI (uMRSI) combined with FID-encoded Iterative Decomposition with Echo Asymmetry and Least-squares estimation (IDEAL), a least-squares chemical shift separation framework using uniformly sampled FID points as temporal encodings at 3.0 T. The proposed uMRSI used only 128 FID points shortening TR to 80 ms (vs. 470 ms for conventional FID-MRSI [cMRSI]), yielding 10.58× shorter acquisition for water-referenced quantification. To reduce bias in major singlet quantification from nonnegligible spectral contributions, the FID-encoded IDEAL signal model also included glutamine-glutamate complex (Glx) and myo-inositol (mI) as singlet surrogates, in addition to tCho, tCr, and tNAA. Unsuppressed water and fat were suppressed using an orthogonality-based subspace-projection framework. Macromolecule (MM) basis functions were parameterized from metabolite-nulled data and incorporated into a two-stage FID-encoded IDEAL scheme that jointly estimated metabolite amplitudes, MM contributions, and local field offset. Phantom validation used healthy adult-range metabolite concentrations with added simulated MM; in vivo evaluation (n = 15) compared uMRSI + FID-encoded IDEAL with cMRSI + LCModel. The theoretical number of signal averages (NSA) approached the maximum for the five modeled metabolites. In phantom, major singlets showed low mean absolute percentage error (MAPE; 12.73% ± 3.76%) and Cramér-Rao lower bound (CRLB; 5.50% ± 1.80%), despite visually noticeable frequency-domain residuals. In vivo, tCho/tCr/tNAA maps showed moderate agreement versus cMRSI + LCModel (structural similarity index measure [SSIM] = 0.71/0.82/0.87; r = 0.67/0.69/0.62 [all p < 0.001]) with small Bland-Altman biases (-0.42/-1.20/-1.98 mM). The proposed uMRSI + FID-encoded IDEAL has potential for further accelerated, water-referenced absolute mapping of the major brain singlets at 3.0 T.
Authors
- Ji‐hoon Kim (ORCID: https://orcid.org/0000-0002-6349-6950)
- Hyeonjin Kim (ORCID: https://orcid.org/0000-0003-2693-9983)
- Joon Hwan Jang (ORCID: https://orcid.org/0000-0001-8934-6494)
Institutions
- Seoul National University (KR)
- Seoul National University of Education (KR)
- New Generation University College (ET)
- Seoul National University Hospital (KR)
- National Medical Center (KR)
Publication Details
- Journal
- NMR in Biomedicine
- Published
- 2026-11-17
- DOI
- https://doi.org/10.1002/nbm.70436
- Primary Topic
- Advanced MRI Techniques and Applications
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Research Foundation of Korea
- Seoul National University Hospital