ViTab Transformer Framework for Predicting Induced Electric Field and Focality in Transcranial Magnetic Stimulation.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Transcranial magnetic stimulation (TMS) is a non-invasive technique for neurological diseases.
• Current prediction models for TMS efficacy use limited input parameters, hindering accuracy.
• Electromagnetic (EM) simulation software is time-consuming.

Purpose of the Study:

• To develop a deep learning-based prediction model for enhanced TMS efficacy.
• To overcome limitations of existing models by incorporating more input parameters.
• To create a faster alternative to EM simulation software for TMS treatment planning.

Main Methods:

• Developed a transformer-based prediction model named ViTab transformer.
• Integrated a vision transformer for image data (MRI) and a tab transformer for tabular data (coil parameters, tissue conductivity, etc.).
• Predicted electric field (E-max), stimulation focality (S-half), and stimulation volume (V-half).

Main Results:

• Achieved high prediction accuracy with R2 scores of 0.97 for E-max, 0.87 for V-half, and 0.90 for S-half.
• The ViTab transformer demonstrated superior accuracy compared to existing state-of-the-art methods.
• Significantly reduced computational time compared to traditional EM simulations.

Conclusions:

• The ViTab transformer model accurately predicts key TMS parameters, offering a significant advancement.
• This model can assist neuroscientists and neurosurgeons in optimizing TMS treatments.
• The combination of speed and accuracy positions ViTab transformer for future clinical applications in neurology.