Deep Learning-Based Development of Personalized Human Head Model With Non-Uniform Conductivity for Brain Stimulation

Area of Science:

• Neuroscience
• Medical Imaging
• Computational Modeling

Background:

• Transcranial magnetic stimulation (TMS) is crucial for brain research and diagnosing neurological disorders.
• Personalized TMS requires accurate individual head models, typically generated through time-consuming MRI segmentation.
• Current models often use uniform conductivity for tissues, which is an oversimplification.

Purpose of the Study:

• To develop a fast and automatic method for estimating human head electrical conductivity for volume conductor models.
• To eliminate the need for anatomical segmentation in head modeling for TMS.
• To enable more accurate, personalized brain stimulation by incorporating position-dependent conductivity.

Main Methods:

• A convolutional neural network (CNN) was designed to estimate electrical conductivity from T1- and T2-weighted MRI scans.
• The CNN estimates conductivity based on water content derived from MRI intensity values.
• This approach bypasses traditional, error-prone tissue segmentation.

Main Results:

• The proposed method provides fast and automatic estimation of personalized head electrical conductivity.
• It successfully avoids the need for manual or automated anatomical segmentation.
• Resulting electric field distributions in the brain are comparable to, yet smoother than, those from conventional methods.

Conclusions:

• This novel AI-driven approach offers a more efficient and potentially more accurate way to create personalized head models for TMS.
• By leveraging MRI intensity for conductivity estimation, it overcomes limitations of segmentation and uniform conductivity assumptions.
• The method holds promise for advancing targeted neurostimulation and neurological disorder diagnosis.