Fiber Orientation Modeling: a Method to Improve Quantitation of Intramyocellular Lipids in Human Subjects at 7 Tesla

Background: Increased intramyocellular lipid (IMCL) content in skeletal muscle has been suggested to be a biomarker for insulin resistance. As a noninvasive method of estimating IMCL, 1H MR spectroscopy of muscle fat has been a popular method for measuring the concentration of IMCLs, a goal highly desirable for research in the pathogenesis of type 2 diabetes. Extramyocellular lipids (EMCL) are often considered to be deposited along strands that are parallel to Bo (the applied field) whereas IMCL are assumed to be spherical droplets in the muscle cells’ cytoplasm. Resolution between IMCL and EMCL signals mainly results from the angle-dependant bulk susceptibility of the 2 geometric structures. However, IMCL signal is usually contaminated by a broad and asymmetrical EMCL . Conventional fitting methods usually assume that both the IMCL and EMCL signals to be symmetrical, represented by a single Lorentzian, Gaussian or Voigt (hybrid lineshape between Gaussian and Lorentzian) lineshapes. However, significant asymmetry in the resonance assigned to the methylene protons (-CH2-)n in extramyocellular lipids (EMCL) interfered with fitting the spectra. In this work, we explore another approach, named Fiber Orientation Modeling (FOM) by using the bulk susceptibility effect theory to accurately assess the lineshape of EMCL.

Methods: The distribution of EMCL strand orientation at any angle from 0 degrees to 90 degrees relative to Bo was described by a Gaussian function, centered at a specific angle and a width representing a dispersion of EMCL strands. The chemical shift OMEGA from each strand was translated by the well-known orientation dependence interaction OMEGA = 3cos2THETA-1, where THETA is the angle between EMCL and the applied field. As the result, the location and amplitude of individual curves representing each strand could be derived. Depend on the aforementioned Gaussian distribution, the combination of these individual curves generated a unique EMCL shape. In this work, spectral simulations were generated using muscle fiber orientation reported previously. The phantom experiment with a fat cylinder (representing EMCL) submerged in Intralipid(TM) solution (representing IMCL) was also performed to determine the maximal shift at 0 degrees and 90 degrees. Under IRB approval, single voxel and chemical shift images were acquired from soleus and gastrocnemius muscle of healthy human subjects at 7T (Phillips Medical system, Cleveland, Ohio). All the spectra were fitted with the hybrid Voigt lineshape and the experimental lineshape.

Results: In simulated spectra with dominant angle of 00 to the applied field and little dispersion, fitting with the Voigt lineshape accurately determined IMCL/EMCL ratio over a range of different linewidths. Increasing dispersion and central angle caused overestimation of IMCL/EMCL ratios, up to three-fold when fitted with the Voigt lineshape. The error was substantially reduced using our method. The improvement is also observed in phantom spectra and human spectra. Estimates of [IMCL]/[EMCL] were significantly improved by including variations in fiber orientation in the lineshape analysis (fiber orientation modeling, FOM). Calculated soleus [IMCL] using FOM, 4.43 ± 2.32 mmol/kg wet weight, was lower compared to most previous reports in soleus. The average orientation of EMCL was calculated to be 35 degrees relative to Bo with a dispersion width of 24 degrees

Conclusion: Since prominent asymmetrical EMCL signal tends to contaminate into IMCL region, this interaction results in the amplitude-dependence of IMCL signal on the average orientation and dispersion of EMCL. As the result, the use of symmetrical lineshape tends to overestimate the IMCL signal if all strands of EMCL are not parallel to Bo and one another. By accounting for the angular dispersion& orientation, the fit would improve both the residual and the IMCL estimate.