Inter-subject receive-template substitution for quantitative brain water content mapping at 3 T: a leave-one-out evaluation

To evaluate a configuration-specific inter-subject receive template as an alternative to participant-specific receive referencing within the evaluated 3 T scanner configuration.

Datasets from twenty-five healthy subjects were analyzed using a strict leave-one-participant-out method. Each degree-3 template used the other 24 participants. Arm E used a participant-specific effective relative receive divisor; Arm T used the held-out template. Both arms underwent independent, identically defined residual calibration, including a local intercept (C₀), its spatially regularized field (μ), and the mean-normalized form used downstream (μ). This was followed by ventricular-CSF calibration. Primary outcomes were participant-level median quantitative water content (qWC) in white and gray matter; whole brain was secondary. Paired bias, mean absolute error (MAE), Bland-Altman limits, Pearson r, and Lin concordance correlation coefficient (CCC) were calculated.

Mean relative receive-template mismatch was 2.273% (standard deviation (SD), 0.568%; range, 1.431%-3.615%). Mean Arm T minus Arm E differences were -0.04 percentage units (p.u.; dimensionless water fraction relative to pure water = 100 p.u.) in white matter and -0.17 p.u. in gray matter; MAE was 0.17 and 0.22 p.u., and CCC was 0.991 and 0.989, respectively. Whole-brain bias was -0.12 p.u. (MAE, 0.19 p.u.; CCC, 0.993). Bland-Altman widths were below 1 p.u. Exploratory associations between global mismatch and qWC disagreement were weak; all 95% participant-bootstrap confidence intervals spanned zero.

Under symmetric arm-specific calibration, the configuration-specific template produced qWC estimates in close agreement with the participant-specific explicit reference strategy. This internal matched-configuration result supports technical feasibility but does not establish formal equivalence, interchangeability, or external generalization.