Repeatability and reproducibility of joint T1-T2 transient-state relaxometry across multiple vendors and implementations at 3T in phantom and human brain

Abstract

Transient-state relaxometry (TSR) enables rapid estimation of T1 and T2 relaxation times. To support its broader adoption in multi-center studies, it is essential to assess the consistency of its implementation across different MRI vendors. This work aimed to assess accuracy, repeatability, and inter-vendor reproducibility of jointly measured T1 and T2 maps based on TSR at 3T. To achieve this goal, a phantom and five volunteers were scanned in a traveling-brain study at four 3T MRI systems from three manufacturers. In the phantom study, Bland-Altman analysis and coefficients of variation (CV) were used to assess accuracy, and repeatability and reproducibility, respectively. Subsequently, in-vivo inter-site variability was evaluated via ANOVA and by computing voxelwise CVs and biases associated with sites were measured via a general linear model (GLM). Excellent accuracy, repeatability, and reproducibility were obtained for the phantom. In-vivo, we found excellent repeatability (CV < 4.5%) and generally good inter-site and inter-vendor reproducibility, though significant variability was found across different TSR implementations. The GLM analysis revealed site-related biases of approximately 100 ms for T1 and 2 ms for T2 in solid brain tissues. These differences may be attributable to different magnetization transfer effects and residual B1+ inhomogeneities due to imperfect calibration. Our findings demonstrate that the bias introduced by the use of different TSR implementations needs to be considered carefully in order to perform in-vivo multi-center studies.