Optimizing synthetic hematocrit for cardiac MRI: a multivariable model calibrated with standardized venous sampling.

Cardiac magnetic resonance (CMR) is a key modality for non-invasive extracellular volume (ECV) quantification, typically requiring venous hematocrit (Hct). However, fluctuations in Hct can affect accuracy. This study aimed to develop a synthetic Hct model integrating age, sex, and pre- and post-contrast blood pool T1 values, calibrated against an optimized venous Hct reference. Cardiac MRI data from 253 patients, including those with various pathologies and normal findings, were retrospectively analyzed for ECV calculations. To minimize postural fluctuations, venous Hct samples were taken after the patients remained in a stable supine position for at least 15 min, just before contrast injection. Hct values were derived from pre-contrast and post-contrast T1 mapping sequences. A comprehensive Hct model was developed by integrating age, sex, and both pre- and post-contrast blood pool T1 values. Unlike previous studies, these variables were evaluated collectively, and the model was calibrated using an optimized venous Hct reference. The performance of the derived model at high ECV values was further evaluated in the hypertrophic cardiomyopathy (HCM) subcohort. The Hct model showed strong agreement with measured ECV (bias: 0.31%, RMSE: 1.25, CCC: 0.949). In the HCM subcohort, it maintained high correlation (r = 0.98) with a clinically acceptable bias of - 1.16%. Compared to sex-specific formulas by Chen et al., the model demonstrated improved performance (R² = 0.56) with lower variability. This standardized synthetic Hct model enables accurate, non-invasive ECV estimation without blood sampling and demonstrates superior performance, especially in patients with elevated ECV values.