Related Experiment Video
Updated: Nov 3, 2025
09:30
Localizing Function-specific Targets for Transcranial Magnetic Stimulation in the Absence of Navigation Equipment
Published on: May 23, 2025
937
Automatic cortical target point localisation in MRI for transcranial magnetic stimulation via a multi-resolution
John S H Baxter1, Quoc Anh Bui2, Ehouarn Maguet2
1Laboratoire Traitement du Signal et de l'Image (LTSI - INSERM UMR 1099), Université de Rennes 1, Rennes, France. jbaxter@univ-rennes1.fr.
Summary
A new convolutional neural network (CNN) achieves human-level performance in localizing brain targets for transcranial magnetic stimulation (TMS), significantly speeding up treatment planning. This AI-driven approach enhances accuracy and efficiency for TMS therapy, especially in resource-limited settings.
Area of Science:
- Neuroscience
- Medical Imaging
- Artificial Intelligence
Background:
- Transcranial magnetic stimulation (TMS) is an emerging therapy for neurological and psychiatric conditions, relying on precise cortical target localization.
- Current methods, including manual MRI-based targeting and atlas-based approaches, are time-consuming and may not fully account for individual anatomical variations.
Purpose of the Study:
- To develop and evaluate a novel, efficient, and accurate method for automatic brain target point localization in MR images for TMS applications.
- To address the limitations of existing methods in terms of speed, computational resources, and handling patient-specific anatomy.
Main Methods:
- A multi-resolution convolutional neural network (CNN) was proposed for localizing predefined points in MR images at varying resolutions.
- The CNN approach was designed for speed, memory efficiency, and the ability to accommodate significant anatomical variability between patients.
Main Results:
- The proposed CNN achieved an accuracy of [Formula: see text] mm, outperforming deformable registration ([Formula: see text] mm).
- The CNN demonstrated human-level performance, with no statistically significant difference compared to a human expert ([Formula: see text] mm).
- The CNN significantly outperformed registration-based methods for most treatment points.
Conclusions:
- The developed CNN offers human-level performance for automated TMS target localization, reducing planning time from hours to seconds.
- This automated approach improves efficiency and accuracy in TMS planning, making it suitable for high-throughput clinical settings and centers with limited computational power.
