SimNIBS-Based Head Model Dataset for Predicting SAR and Electric Field Distributions in Temporal Interference Stimulation

This dataset contains simulation-based measurements of Specific Absorption Rate (SAR) and Electric Field (E-field) distributions for temporal interference deep brain stimulation (tTI-DBS). It was generated using the SimNIBS 4.5 pipeline across five anatomically realistic head models (m2msub01, m2msub09, m2msub10, m2msub12, and m2msub15).Each model was simulated across frequencies ranging from 2000\u20132020 Hz, with bilateral electrode montages targeting the subthalamic nuclei (STN). The dataset integrates SAR statistics, electrode geometry, distance measures, frequency parameters, and demographic attributes to support both regression (SAR prediction) and classification (safety estimation) machine learning tasks.Each entry represents a unique combination of subject, frequency, and electrode configuration. Safety labels were derived using a threshold of 0.28 W\/kg, corresponding to 14% of the IEEE guideline for localized SAR.The dataset enables reproducible and explainable modeling of brain heating during non-invasive stimulation, contributing to safety assessment and optimization of stimulation protocols in computational neuroscience.