Brain temperature plays a vital role in brain health, function, and recovery from injury. Maintaining brain temperature homeostasis involves a balance between metabolic heat generation and heat dissipation via arterial blood flow. Disruptions in this balance due to injury or disease can lead to significant changes in brain temperature, impacting patient outcomes. Studies have shown a strong correlation between increased brain temperature after injuries like traumatic brain injury (TBI) and stroke, and worse patient outcomes. Even small temperature increases (1°C) can correlate with increased tissue damage and mortality. Conversely, therapeutic hypothermia (systemic cooling) can mitigate brain damage and improve recovery. Despite its importance, brain temperature is understudied due to the lack of readily available non-invasive measurement methods. Body temperature is often used as a surrogate, but significant discrepancies between brain and body temperatures have been observed, especially after injury, highlighting the need for accurate brain thermometry. Various experimental methods for brain thermometry have been developed using temperature-sensitive magnetic resonance (MR) parameters, including proton resonance frequency (PRF)-based MR thermometry, which shows promise for non-invasive in vivo temperature mapping. However, these methods require further development and validation before widespread clinical adoption. Biophysical models offer a valuable alternative, but many lack adherence to first principles of energy and mass conservation and fail to account for individual variations in brain anatomy and physiology. The current study aims to address these limitations by developing a biophysical model based on fundamental conservation laws that incorporates individual MRI data to generate personalized brain temperature predictions.

Existing literature highlights the significant impact of brain temperature on neurological outcomes following injury or disease. Studies have demonstrated a clear link between elevated brain temperature and increased tissue damage and mortality after TBI and stroke. Conversely, therapeutic hypothermia has shown promise in mitigating injury and improving recovery. However, the lack of non-invasive and readily available methods for measuring brain temperature poses a significant challenge. While various experimental techniques, such as PRF-based MR thermometry, exist, they often lack the accuracy, resolution, and clinical practicality required for routine use. Prior biophysical models attempted to address this gap but often fell short due to simplifications that ignored fundamental physical laws or neglected patient-specific anatomical variations. This lack of accurate, personalized temperature predictions underscores the need for more sophisticated models that can capture the complexities of brain thermoregulation.

This study developed a biophysical model of brain temperature based on first principles of mass and energy conservation. The model incorporates three heat transfer domains (arteries, veins, and tissue) and three modes of energy transfer (conduction, convection, and advection). The key steps involved in the model's construction are: 1. **Individual subject data acquisition:** MRI data (MRA, MRV, T1-weighted images) were collected from three healthy subjects to obtain individual arterial, venous, and tissue structure maps. 2. **Vasculature augmentation:** The MR-derived vasculature, consisting of major arteries and veins, was augmented with smaller vessels using a rapidly exploring random tree (RRT) algorithm. The number of RRT iterations was optimized to obtain physiologically realistic cerebral blood flow (CBF) values and distributions. 3. **Flow rate determination:** Flow rates were calculated for each vessel segment and across tissue using quasi-steady-state mass conservation, considering total blood supply to the brain and hydrodynamic resistance. 4. **Metabolic rate calculation:** Local metabolic rates of heat generation were determined on a voxel-wise basis using probability-weighted imaging for gray and white matter densities. 5. **Governing equations:** A system of steady-state governing equations, based on local energy conservation, was formulated for each domain (arteries, tissue, veins). These equations capture the interactions between domains through conductive, convective, and advective heat transfer. 6. **Model solution:** The system of equations was solved numerically to obtain a 3D temperature distribution across the brain. 7. **Model validation:** The model's predictions were compared with experimental temperature measurements obtained using whole-brain MR chemical shift thermometry. The comparisons focused on voxel-wise temperature distributions and overall mean temperatures. A sensitivity analysis was performed to evaluate the effect of tissue voxel size on the accuracy of temperature predictions.

The study successfully generated subject-specific brain temperature maps using a biophysical model that rigorously incorporates individual MRI data and adheres to fundamental conservation laws. The key findings include: 1. **Subject-specific temperature variations:** Significant spatial variations in brain temperature were observed across the three subjects, highlighting the importance of personalized modeling. These variations reflected individual differences in tissue structure, vessel architecture, and blood flow. 2. **Impact of vessel augmentation:** The RRT algorithm effectively augmented the MR-derived vasculature, resulting in more realistic CBF distributions and temperature predictions. Increasing the number of RRT iterations led to decreased mean CBF, more uniform temperature distributions, and smaller vessel diameters, aligning better with physiological expectations. 3. **Voxel size sensitivity:** The model's sensitivity to tissue voxel size was investigated, with results suggesting that a voxel size smaller than or equal to the maximum vessel segment length is sufficient to accurately capture spatial temperature variations. Larger voxel sizes resulted in underestimation of thermal variations due to a coarser CBF distribution. 4. **Model-experiment agreement:** Strong agreement was found between model-predicted and MR-measured brain temperatures. Most voxels showed temperature differences within a threshold of ±0.8°C. Discrepancies observed in frontal lobe regions were attributed to limitations in the experimental MR data acquisition. 5. **Comparison with a generic model:** The study compared subject-specific temperature predictions with those from a generic model based on atlas data. The subject-specific model demonstrated better capture of local temperature variations than the generic model which failed to predict the significant localized temperature differences due to individual anatomy. This highlights the superiority of the proposed personalized approach. The results demonstrated that both model-predicted and MR-measured brain temperatures were higher than axillary body temperature, with mean model-predicted temperatures being ≤0.5 °C higher than core body temperature.

The findings demonstrate the feasibility of creating personalized brain temperature predictions using a biophysical model driven by individual MRI data and grounded in the principles of energy and mass conservation. The significant subject-specific temperature variations observed underscore the limitations of generic brain models and the need for personalized approaches in predicting and managing brain temperature, particularly after injury or disease. The strong agreement between model predictions and experimental MR temperature measurements validates the accuracy of the model and supports the hypothesis that individual anatomical variations significantly influence brain temperature distributions. The optimization of the RRT algorithm for vessel augmentation and the sensitivity analysis for tissue voxel size highlight the importance of considering physiological and computational constraints during model development. These results have implications for the improvement of therapeutic interventions such as hypothermia, optimizing treatment based on individual patient characteristics. The limitations of the current model, such as its steady-state nature and the simplifications made in the representation of the capillary network and inter-hemispheric heat exchange, point to future research directions.

This study presents a novel biophysical model capable of generating subject-specific predictions of brain temperature distributions. The model leverages individual MRI data and adheres to first principles of energy and mass conservation, resulting in accurate and personalized predictions. The strong validation against experimental data and superior performance compared to generic models strongly suggests a path forward for personalized thermal management approaches in clinical settings. Future research should focus on incorporating dynamic aspects of blood flow and metabolism, refining the capillary network representation, and improving the accuracy and resolution of whole-brain MR thermometry to further enhance model fidelity.

The current study has several limitations. The model currently assumes steady-state conditions and does not capture the dynamic changes in brain temperature that occur in response to physiological events. The representation of the capillary network was simplified, and inter-hemispheric heat exchange was not fully accounted for. Furthermore, the study's sample size was small, limiting the generalizability of the findings. The MR thermometry technique used also has limitations in terms of spatial resolution and susceptibility to artifacts, particularly near the sinuses, affecting the accuracy of the experimental temperature measurements.