Intrinsic Divergence, Repeatability, and Distributional Fingerprints of VFA, ME-SE, MDME, and MRF: A Comparative Evaluation of Quantitative T<sub>1</sub>/T<sub>2</sub> Relaxometry in Phantom and Human Brain at 3 T.
Authors
Affiliations (6)
Affiliations (6)
- Department of Magnetic Resonance Imaging, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, China.
- Department of Radiology, Zhecheng County Hospital of Traditional Chinese and Western Medicine, Shangqiu, China.
- Department of Radiology, Charité - Universitätsmedizin Berlin, Berlin, Germany; MR Research Collaboration, Siemens Healthineers, Beijing, China.
- MR Research Collaboration, Siemens Healthineers, Beijing, China.
- Department of Magnetic Resonance Imaging, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, China.. Electronic address: [email protected].
- Department of Magnetic Resonance Imaging, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, China.. Electronic address: [email protected].
Abstract
Quantitative MRI (qMRI) relaxometry provides non-invasive biomarkers of brain microstructure, yet cross-method inconsistencies continue to hinder reliable comparison across studies and sites. To facilitate the clinical standardization of brain qMRI, this work systematically evaluated the accuracy and repeatability of three clinical brain T<sub>1</sub>/T<sub>2</sub> relaxometry implementations under harmonized 3 T conditions: conventional variable flip-angle and multi-echo spin-echo (VFA/ME-SE), multi-dynamic multi-echo (MDME), and magnetic resonance fingerprinting (MRF). A standardized ISMRM/NIST system phantom and two healthy volunteer cohorts were examined. The phantom experiment quantified accuracy and bias; a multi-site "traveling brain" cohort (n = 12) assessed inter-scanner repeatability; and a single-site population cohort (n = 38) characterized distributional "fingerprints" and biological sensitivity using a high-precision AI-based segmentation pipeline. In phantom validation, MRF achieved the highest accuracy and stability across the physiological range, whereas ME-SE exhibited precision loss at long T<sub>2</sub>. In vivo, all methods demonstrated excellent inter-site repeatability with coefficients of variation (CVs) below 5%, where MRF achieved the highest stability for T<sub>1</sub> (CV = 0.61%) and MDME yielded the highest stability for T<sub>2</sub> (CV = 0.42%). However, substantial intrinsic discrepancies persisted: relative to MRF, VFA systematically overestimated T<sub>1</sub> (particularly in deep gray matter), while MDME underestimated T<sub>1</sub>. For T<sub>2</sub>, a fundamental baseline shift was observed, with ME-SE and MDME yielding values nearly double those of MRF in iron-rich regions. Supplementary investigations confirmed these offsets arise from proprietary reconstruction and signal encoding differences (transient-state vs. steady-state) rather than simple protocol constraints. Biologically, the optimized segmentation pipeline enhanced sensitivity to age-related trends, revealing significant T<sub>1</sub> shortening across all methods, while VFA alone exhibited significant sex-bias confounding not observed in MRF or MDME. These findings provide quantitative benchmarks and practical guidance for standardizing brain qMRI relaxometry across acquisition methods, scanners, and research sites.