Abstract
PURPOSE: To evaluate the repeatability of MRI-derived cell size and density measurements in the healthy human liver using MR cytometry, a non-invasive technique for quantifying tissue microstructure, METHODS: Eight healthy subjects underwent two identical MR cytometry scans (2-7 days apart) on a Phillips 3 T scanner. Diffusion data were acquired using pulsed gradient spin echo (PGSE) and oscillating gradient spin echo (OGSE) sequences at multiple diffusion times. Bayes factor model selection compared different MR cytometry signal models, such as IMPULSED, which ignores water exchange between intra and extracellular compartments, and JOINT, which accounts for it. Repeatability of MRI-derived parameters were assessed via whole-volume and multi-size ROI analyses using Bland-Altman plots and correlation coefficients. RESULTS: The JOINT model was decisively preferred (>90 % of voxels). High repeatability was observed for d, v(in), ADC(ex), and cell density in whole-liver and larger ROIs, with correlations up to r = 0.96. ROI size significantly improved consistency. Coefficients of variation for cell density decreased from 72 % (single voxel) to 28 % (7 × 7 voxels). The exchange rate k(in) could not be reliably estimated. CONCLUSION: ROI-based MR cytometry provides repeatable measurements of mean cell size and density, supporting its potential for clinical applications such as assessing steatohepatitis and detecting hepatocellular carcinoma. Protocol optimization to improve SNR is needed for reliable quantification of water exchange.