Prediction of Stimulation Strength of Transcranial Magnetic Stimulation in the Brain with Deep Encoder-Decoder Convolutional Neural Network

Transcranial magnetic stimulation (TMS) is a non-invasive, effective, and safe neuromodulation technique to treat neurological and psychiatric disorders. However, it is difficult to precisely assess whether crucial brain areas have received the appropriate degree of generated electric field due to the complexity and variety of the brain's composition and structure. Electric field distribution may be calculated numerically using finite element analysis (FEA). These techniques take a lot of time and require extremely high computing resources. In this study, we created a deep convolutional neural network (DCNN) encoder-decoder model to predict induced electric fields from T1- and T2-weighted MRI-based anatomical slices in real time. To test resting motor thresholds, we gathered 11 healthy subjects and applied TMS on the primary motor cortex. SimNIBS pipeline was used to create head models from the subjects' MRI scans. For each subject, the overall size of the head models was scaled to 20 new size scales, resulting in a total of 231 head models. The quantity of input data representing various head model sizes was scaled up. The induced electric fields were calculated using Sim4Life, a FEA program, and utilized as the DCNN training data. Peak signal to noise ratios for the training and testing sets of data for the trained network were 32.83 dB and 28.01 dB, respectively. Our model's primary contribution is its capability to forecast the induced electric fields in real-time, which enables us to precisely and effectively estimate the required TMS intensity in the targeted brain areas.