Surrogate Model Supported Optimization of high-definition Temporal Interference Stimulation

Temporal interference stimulation is a non-invasive neuromodulation technique using two or more electrodes to generate an electrical field, targeting deep brain structures. High-definition temporal interference stimulation enhances the focality and intensity of stimulation in specific regions. However, optimizing electrode positions remains challenging in TI stimulation. Typically, predefined discrete electrode positions are used, resulting in suboptimal outcomes due to the limited exploration of continuous positions. We address this by optimizing electrode placement freely on the head surface, leveraging high-fidelity simulations of the electrical field. To manage computational complexity, we propose using a neural network to approximate the mapping from continuous electrode positions to stimulation intensity in the target region. This surrogate model is used in the optimization process, significantly reducing computing time. Our results show that optimization based on the surrogate model for continuous electrode positions enhances stimulation performance, balancing intensity and focality.