Sarah Keller1, Avneesh Chhabra2, Shaheen Ahmed3, Anne C Kim4, Jonathan M Chia5, Jin Yamamura6, Zhiyue J Wang7. 1. University Medical Center Hamburg-Eppendorf, Hamburg, Germany. Electronic address: s.keller@uke.de. 2. University of Texas Southwestern Medical Center, Dallas, TX, USA. Electronic address: Avneesh.Chhabra@UTSouthwestern.edu. 3. University of Texas Southwestern Medical Center, Dallas, TX, USA. Electronic address: Shaheen.Ahmed@utdallas.edu. 4. The Permanente Medical Group, San Francisco, CA, USA. Electronic address: Anne.C.Kim@kp.org. 5. Clinical Science, Philips Healthcare, Cleveland, OH, USA. Electronic address: jonathan.m.chia@philips.com. 6. University Medical Center Hamburg-Eppendorf, Hamburg, Germany. Electronic address: j.yamamura@uke.de. 7. University of Texas Southwestern Medical Center, Dallas, TX, USA; Children's Medical Center, Dallas, TX, USA. Electronic address: jerry.wang@childrens.com.
Abstract
OBJECTIVE: Quantitative diffusion tensor imaging (DTI) of skeletal muscles is challenging due to the bias in DTI metrics, such as fractional anisotropy (FA) and mean diffusivity (MD), related to insufficient signal-to-noise ratio (SNR). This study compares the bias of DTI metrics in skeletal muscles via pixel-based and region-of-interest (ROI)-based analysis. METHODS: DTI of the thigh muscles was conducted on a 3.0-T system in N = 11 volunteers using a fat-suppressed single-shot spin-echo echo planar imaging (SS SE-EPI) sequence with eight repetitions (number of signal averages (NSA) = 4 or 8 for each repeat). The SNR was calculated for different NSAs and estimated for the composite images combining all data (effective NSA = 48) as standard reference. The bias of MD and FA derived by pixel-based and ROI-based quantification were compared at different NSAs. An "intra-ROI diffusion direction dispersion angle (IRDDDA)" was calculated to assess the uniformity of diffusion within the ROI. RESULTS: Using our standard reference image with NSA = 48, the ROI-based and pixel-based measurements agreed for FA and MD. Larger disagreements were observed for the pixel-based quantification at NSA = 4. MD was less sensitive than FA to the noise level. The IRDDDA decreased with higher NSA. At NSA = 4, ROI-based FA showed a lower average bias (0.9% vs. 37.4%) and narrower 95% limits of agreement compared to the pixel-based method. CONCLUSION: The ROI-based estimation of FA is less prone to bias than the pixel-based estimations when SNR is low. The IRDDDA can be applied as a quantitative quality measure to assess reliability of ROI-based DTI metrics.
OBJECTIVE: Quantitative diffusion tensor imaging (DTI) of skeletal muscles is challenging due to the bias in DTI metrics, such as fractional anisotropy (FA) and mean diffusivity (MD), related to insufficient signal-to-noise ratio (SNR). This study compares the bias of DTI metrics in skeletal muscles via pixel-based and region-of-interest (ROI)-based analysis. METHODS: DTI of the thigh muscles was conducted on a 3.0-T system in N = 11 volunteers using a fat-suppressed single-shot spin-echo echo planar imaging (SS SE-EPI) sequence with eight repetitions (number of signal averages (NSA) = 4 or 8 for each repeat). The SNR was calculated for different NSAs and estimated for the composite images combining all data (effective NSA = 48) as standard reference. The bias of MD and FA derived by pixel-based and ROI-based quantification were compared at different NSAs. An "intra-ROI diffusion direction dispersion angle (IRDDDA)" was calculated to assess the uniformity of diffusion within the ROI. RESULTS: Using our standard reference image with NSA = 48, the ROI-based and pixel-based measurements agreed for FA and MD. Larger disagreements were observed for the pixel-based quantification at NSA = 4. MD was less sensitive than FA to the noise level. The IRDDDA decreased with higher NSA. At NSA = 4, ROI-based FA showed a lower average bias (0.9% vs. 37.4%) and narrower 95% limits of agreement compared to the pixel-based method. CONCLUSION: The ROI-based estimation of FA is less prone to bias than the pixel-based estimations when SNR is low. The IRDDDA can be applied as a quantitative quality measure to assess reliability of ROI-based DTI metrics.
Authors: Teodoro Martín-Noguerol; Rafael Barousse; Daniel E Wessell; Ignacio Rossi; Antonio Luna Journal: Eur Radiol Date: 2022-05-12 Impact factor: 5.315