Abstract
Transcranial Magnetic Stimulation (TMS) is a non-invasive neuromodulation method that can modulate neural activity by inducing an electric field in the brain. Computational modeling of electric fields is a major tool for personalizing TMS parameters. State-of-the-art modeling techniques use numerical methods such as the finite element method (FEM) that can produce highly accurate simulation results but operate at a high computational cost (several hours per individual). This has resulted in a bottleneck precluding real-time applications. Recent advancements in deep learning (DL) have demonstrated the effectiveness of deep neural networks in processing magnetic resonance (MR) image data effectively. Here, we develop a DL framework to estimate TMS-induced electric fields directly from the anatomical MR image and TMS stimulation parameters. Such a framework would dramatically reduce the computational cost associated with TMS electric field simulations and allow for more effective protocol optimization and the development of real-time applications of TMS. We are constructing a dataset of 100 MR scans from a diverse population demographic (ethnic, gender, age) made available by the Human Connectome Project. We are generating a FEM head model for each and simulating the electric fields for 13 TMS coil orientations and 1206 positions (total of 15,678 coil placements). We train a modified U-NET architecture to predict individual TMS-induced electric fields in the brain based on an input T1w MR scan and stimulation parameters. We characterize the model’s performance according to the computational efficiency and simulation accuracy. The preliminary results demonstrate a steep acceleration of electric field modeling speed to 0.1 seconds per simulation (×97,000 times acceleration over the FEM-based approach). The agreement between the DL-generated and FEM-based electric fields is very high (less than 1.5 mm deviation in the center of gravity). Our findings demonstrate the potential of DL to accelerate the accurate prediction of TMS-induced electric fields.