0415
Mapping of Muscle Strain Rate at Varying Force Levels of Isometric Contraction, with Compressed-Sensing Velocity-Encoded Phase-Contrast MR Imaging.1Radiology, UC San Diego, San Diego, CA, United States, 2Physics, UC San Diego, San Diego, CA, United States, 3Physics, San Diego State University, San Diego, CA, United States
Synopsis
Strain rate (SR) tensor mapping can be conveniently computed from velocity encoded phase contrast (VE-PC) imaging. The study of the variation of strain rate indices with force output (% Maximum Voluntary Contraction (MVC)) can provide additional information similar to stress-strain relationships measured at the whole muscle level. However, such studies have been limited by the long VE-PC sequence time precluding its use at high MVCs. We have developed a compressed sensing VE-PCI technique to enable acquisitions across a range of MVCs. Successful SR mapping for 30-70%MVC on six subjects is reported here.
INTRODUCTION
The ability to image tissue deformations provides a non-invasive way to study muscle kinematics at the voxel level. Strain rate (SR) imaging based on gated velocity encoded phase contrast imaging has been recently established as a viable methodology to extract strain rate parameters that captures tissue deformations1. The variation of the SR indices with force output (% Maximum Voluntary Contraction, MVC) can provide information on stress-strain like relationships. However, such studies have been limited by human fatigue from the long scanning time, since each phase-encoding level has to be gated, precluding its use at high MVCs. We have developed a compressed sensing(CS) VE-PCI technique to enable acquisitions across a range of MVCs. The CS technique combines multiple coils and a kyt-SENSE SPARSE reconstruction to obtain artifact free images in 37 seconds at a given %MVC effort. SR maps were derived from the velocity images to quantify the SR along the fiber direction and the maximum value of SRfiber showed an increase with %MVC.METHODS
Six subjects were recruited after IRB approval and scanned on a 1.5T GE scanner. Imaging protocol included a set of gated VE-PC images obtained during isometric contraction (TE: 7.7ms, TR: 16.4ms, NEX: 2, FA: 20°, slice thickness 5mm, sagittal-oblique orientation, FOV: 30×22.5cm (partial-phase FOV: 0.75), matrix: 256×106 (undersampling factor 4), 4 views/segment, 8 phases, 3D velocity encoding, venc: 10cm/s, 14 repetitions, cycle length 2.9sec). For the multi-coil CS scheme, a ky-t undersampling followed a variable density random undersampling with maximum density at the center of k-space. A two-step ky-t-SENSE-SPARSE CS joint reconstruction (of reference and velocity encoded images) was performed2 using the coil sensitivities with a temporal FFT followed by a Temporal Total Variation as the sparsifying transforms. A CS undersampling factor of 4 was used to acquire dynamic data in 37 seconds.
The lower leg was placed in a plaster cast with an embedded strain sensor and anchored in an 8-channel custom RF coil; real-time visual feedback was provided to the subject. Data sets were obtained for peak forces corresponding to 30-70% MVC for all subjects. Prior to the analysis, phase-contrast images were corrected for phase shading artifacts and denoised using 2D anisotropic diffusion filter1. 2DSR tensor was calculated from the velocity images by taking spatial gradient and then symmetrized. Strain rate eigenvalues (SRfiber, SRin-plane) were obtained from SR tensor through eigenvalue decomposition1.
Quantitative analysis was performed for a ROI placed in medial gastrocnemius muscle (7x7). Position of each voxel inside ROI was tracked in plane across the cycle. Strain rate indices were extracted at the frame corresponding to max SRfiber during contraction part of the cycle.
RESULTS
Fig. 1 and Fig. 2 show the SRfiber and SRin-plane maps respectively at different %MVC effort; the maps correspond to the temporal frame at max SRfiber in the contraction cycle. The quality of the SR maps underlines the efficiency of the CS reconstruction even when the reduction factor is high. Fig. 3 shows the temporal maps of the SR indices as a function of the isometric contraction cycle for one subject; different curves correspond to different %MVCs. The increase in the magnitude of SR values with %MVC can be readily appreciated.DISCUSSION
CS-VEPC decreased scan time by a factor of 4 enabling all subjects to complete the repetitions required for a dynamic scan even at 70% MVC. Strain rate maps are sensitive to noise in the velocity maps and there is a loss of SNR in CS-VEPC as fewer phase encodes are collected for undersampling. However, the quality of the SR maps acquired with the CS-VEPC show that they have adequate SNR for quantitative analysis. Strain rate along the fiber direction increased with %MVC showing the increased contraction of the muscle fiber in order to produce more force. Since the contraction in the fiber direction is compensated by an expansion in the fiber cross-section, it stands to reason that an increased contraction along the muscle fiber (SRfiber) will be accompanied by an increase in the fiber cross-section(SRin-plane). This is seen as an increase in SRin-plane with % MVC.CONCLUSIONS
This the first report of the variation of SR indices with %MVC (upto 70%MVC) and this was enabled by the compressed sensing VE-PC which decreased scan times to ~38 seconds at a given MVC. The magnitude of the SR indices increased with % MVC showing that the proposed technique can be used to explore tissue deformations as a function of %MVC up to much higher %MVC than without CS.Acknowledgements
This work was supported by the National Institute of Arthritis and Musculoskeletal and Skin Diseases Grant R01AG056999-01References
[1] Malis V, Sinha U, Csapo R, Narici M, Sinha S. Magn Reson Med. 2017;doi: 10.1002/mrm.26759.
[2] Kim D, Dyvorne HA, Otazo R, et al. Magn Reson Med. 2012; 67: 1054-1064.
Figures
