Parametric imaging methods (e.g., T ₁ relaxation time mapping) have been shown to be more reproducible across time and vendors than weighted (e.g., T ₁ -weighted) images. The purpose of this work was to more extensively evaluate the validity of this assertion.
Seven volunteers underwent twice-repeated acquisitions of variable flip-angle T ₁ mapping, including B ₁ ⁺ calibration, on a 3T Philips Achieva and 3T Siemens Trio scanner. Intra-scanner and inter-vendor T ₁ variability were calculated. To determine T ₁ reproducibility levels in longitudinal settings, or after changing hardware or software, four additional data sets were acquired from two of the participants; one participant was scanned on a different 3T Siemens Trio scanner and another on the same 3T Philips Achieva scanner but after a software upgrade.
Intra-scanner variability of voxel-wise T ₁ values was consistent between the two vendors, averaging 0.7/0.7/1.3/1.4% in white matter/cortical gray matter/subcortical gray matter/cerebellum, respectively. We observed, however, a systematic bias between the two vendors of https://doi.org/10.0/7.8/8.6/10.0%, respectively. The T ₁ bias across two scanners of the same model was greater than intra-scanner variability, although still only at 1.4/1.0/1.9/2.3%, respectively. A greater bias was identified for data sets acquired before/after software upgrade in white matter/cortical gray matter (3.6/2.7%) whereas variability in subcortical gray matter/cerebellum was comparable (1.7/1.9%).
We established intra- and inter-vendor reproducibility levels for a widely used T ₁ mapping protocol. We anticipate that these results will guide the design of multi-center studies, particularly those encompassing multiple vendors. Furthermore, this baseline level of reproducibility should be established or surpassed during the piloting phase of such studies.