First-principles multiscale modeling of cerebral hemodynamics enables personalized predictions of human brain temperature.

Computational bioheat modeling offers a powerful approach for characterizing brain temperature when direct measurements are limited. We developed a multiscale biophysical model with conservation of mass, momentum, and energy across all spatial scales to predict human brain temperature and compared the temperature maps with a previously developed model without global momentum conservation and in vivo magnetic resonance (MR) thermometry. Subject-specific brain anatomy obtained from MR images was incorporated into bioheat equations, integrating physiological parameters such as metabolic heat and cerebral blood flow. A novel approach using "transitional" microvasculature was introduced to bridge macro- and micro-scale domains, enabling momentum conservation and accurate prediction of cerebral hemodynamics. Across 30 healthy subjects, voxel-wise mean absolute differences of 0.18-0.36 °C between modeled and experimentally measured temperatures were observed. ROI-based analysis showed significant correlations between model predictions and MR measurements of brain temperature when system-wide momentum is properly conserved, but no correlation was observed for model predictions without momentum conservation. A simulated scenario of middle cerebral artery occlusion highlighted the importance of momentum conservation to mimic realistic physiological changes. The agreement of model predicted temperatures with MR measurements support future clinical applications where brain temperature may serve as a biomarker for neurological disorders.