Apparent Intramyocellular Lipid Index Using Spiral Magnetic Resonance Spectroscopic Imaging at 3 T

Abstract

Abstract Purpose: This work introduces a rapid, high-resolution method based on spiral Magnetic Resonance Spectroscopy Imaging (MRSI) to map the apparent content of intramyocellular (IMCL) and extramyocellular (EMCL) lipids in human skeletal muscle. Methods: Unsuppressed water spiral MRSI data were acquired from healthy volunteers using a dual-tuned 1H/31P transmit/receive coil positioned under the right calf. Frequency registration of spectra was ensured using the Free Induction Decay (FID) modulus approach. By computing the cumulative sum of the normalized spectral amplitude over the 1.09–1.71 ppm range, we derived the Apparent Intramyocellular Lipid Index (AIMLI), which provides an estimate of the relative IMCL contribution within the total lipid signal (IMCL + EMCL) in each voxel. Signal simulations were conducted to optimize the AIMLI’s method to minimize its sensitivity to frequency shifts of the EMCL peak induced by magnetic susceptibility effects. In vivo, AIMLI performance was compared to classical LCModel quantification in muscles of interest (Gastrocnemius Medialis (GM) and Soleus Medialis (SM)) and discussed regarding fibers’ orientation (assessed from SE-EPI diffusion-weighted), and high-resolution water, fat, and fat fraction images (derived from 3D Chemical-Shift-Encoded (CSE) gradient multiecho T1 VIBE Dixon data). Repeatability and reproducibility were systematically evaluated. Additionally, AIMLI was applied in a longitudinal study to monitor metabolic changes during long-term fasting and compared with conventional Single Voxel Spectroscopy (SVS)-LCModel quantification. Results: Simulation confirmed that AIMLI generates maps consistent with its quantitative counterpart and identified the optimal chemical shift for index computation. Both the AIMLI and its quantitative LCModel equivalent revealed significant differences between GM and SM muscles (p < 0.05). A significant positive correlation was observed between AIMLI and LCModel-derived values, while AIMLI demonstrated lower coefficients of variation for repeatability and reproducibility. Group-level AIMLI values demonstrated significant differences across the three time points (p < 0.05). Posthoc comparisons revealed a significant increase in mean AIMLI after fasting (D + 12), followed by a return to baseline at D + 30. These trends were consistent with those observed with SVS-LCModel data, supporting the physiological relevance of AIMLI. Conclusion: AIMLI offers a spatially resolved, rapid, and robust approach to map apparent IMCL content relative to total lipids. Preliminary in vivo maps highlighted higher IMCL content in the slow-twitch SM muscle, consistent with its high proportion of type I fibers, and lower IMCL in the GM. With its short acquisition time and resilience to spectral distortions, AIMLI holds promise for enhancing clinical feasibility in monitoring lipid distribution across physiological and metabolic disorders. These advantages motivate its adoption in future clinical studies targeting skeletal muscle metabolism.

The authors' abstract, as published at the source. Chemical & Biomedical Imaging, 2026 · DOI ↗