Abstract
BACKGROUND: Dixon-based magnetic resonance imaging (MRI) intramuscular proton density fat fraction (PDFF) is a potentially useful imaging biomarker of muscle quality. However, multi-vendor, multi-site reproducibility of intramuscular PDFF quantification, required for large clinical studies, can be strongly dependent on acquisition and processing. The purpose of this study was (I) to develop a 6-point Dixon MRI-based acquisition and processing technique for reproducible multi-vendor, multi-site quantification of thigh intramuscular PDFF; and (II) to evaluate the ability of the technique to detect differences in thigh muscle status between operated vs. non-operated limbs in a multi-site study of patients scanned at 10 years after anterior cruciate ligament reconstruction (ACLR). METHODS: MRI bilateral mid-thigh data acquisition at 3T was harmonized across three sites and two vendors and included high-resolution axial T1-weighted scans and 6-point Dixon scans. Centralized, vendor-independent PDFF quantification was performed and algorithms were evaluated in phantoms to determine the most reproducible approach. A novel image post-processing method was developed to mitigate scaling errors observed on some scanner platforms to improve reproducibility. PDFF measurements in phantoms and control subjects including traveling controls were obtained for assessment of intra-scanner repeatability as well as inter-scanner, inter-vendor, and inter-site reproducibility. Patients from the Multicenter Orthopedic Outcomes Network ACLR cohort were scanned and intramuscular PDFF was compared between thigh muscles of the operated and contralateral limbs. Standard deviation (SD) of PDFF, within-subject SD (wSD), and intraclass correlation coefficient (ICC) were used to characterize repeatability and reproducibility. RESULTS: The proposed scaling correction method improved overall reproducibility in phantoms and traveling controls and was incorporated as part of the Dixon processing pipeline for subsequent analyses. Intra-scanner phantom repeatability ranged between 0.2-0.9% (SD) PDFF (ICC =0.98-1.00), with overall inter-vendor/inter-site reproducibility of 0.7-1.7% (SD) PDFF (ICC =0.97). Control subject repeatability among all scanners and vendors ranged between 0.2-0.8% (wSD) PDFF (ICC =0.95-0.98) with slightly lower inter-site, inter-vendor reproducibility, 0.8-1.2% (wSD) PDFF (ICC =0.92). Intramuscular PDFF was elevated in ACLR vs contralateral thighs for the hamstrings muscle compartment (6.2%±3.5% vs. 5.7%±2.8%, P=0.006), while quadriceps (4.0%±2.0% vs. 4.0%±2.1%, P=0.961) and medial (5.7%±2.5% vs. 5.5%±2.3%, P=0.133) muscle compartments did not show significant PDFF differences. CONCLUSIONS: Reproducible multi-site, multi-vendor intramuscular PDFF measurement is enabled by 6-point Dixon MRI with standardized acquisition and processing. The method is sensitive enough to detect differences in muscle groups between operated and non-operated thighs in the patient population 10 years after ACLR.