Noninvasive prediction of photon fluence rate in targeted brain regions for personalized transcranial photobiomodulation dosage: model development and in silico evaluation

Transcranial photobiomodulation (tPBM) shows promise in delivering beneficial effects to the brain. However, accurately estimating the stimulus energy reaching the targeted brain region remains difficult due to individual differences in anatomy and optical properties. We present a noninvasive method that combines diffuse reflectance spectroscopy with deep learning to predict the fluence rate of the stimulus light. Incorporating tissue layer thicknesses into the model significantly enhances prediction accuracy, reducing errors to approximately 13%, compared to 49% when assuming a constant irradiance at the scalp surface. By eliminating the need for expensive magnetic resonance imaging, our approach offers a scalable solution for optimizing irradiation parameters in future tPBM applications.