Transcranial Magnetic Stimulation (TMS) is a non-invasive technique used to study brain function and to influence neural activity in both research and clinical settings. One important application of TMS is brain mapping, which helps researchers better understand how brain structure and function are connected. For TMS-based brain mapping to be effective, it is crucial to stimulate the right locations in the brain with high precision.
The challenge with current approaches
Modern TMS motor mapping methods use statistical models to describe the relationship between the electric field induced by TMS in the brain and the resulting muscle response, known as the motor evoked potential (MEP). While these approaches represent a clear improvement over traditional techniques, they usually rely on randomly chosen stimulation locations.
This random sampling often leads to redundant measurements, meaning that many stimulation trials provide little new information. As a result, more trials are needed to obtain an accurate motor map, increasing the duration and burden of experiments, especially for patients.
Our team
To tackle these challenges, our team at the Department of Neuroanatomy works hand in hand with outstanding experts across disciplines. We collaborate closely with computer scientists specializing in machine learning and artificial intelligence, as well as with clinicians including neurologists and neuroradiologists.
Our international team brings together leading researchers from Freiburg and Tromsø (Norway), combining complementary expertise to drive innovative, interdisciplinary brain research.
A smarter stimulation strategy
In our study, we introduce an optimization strategy to make TMS-based motor mapping faster and more efficient without compromising accuracy.
Our approach uses a method called farthest point sampling (FPS). In simple terms, FPS is a smart way of deciding where to stimulate next. Instead of choosing locations randomly, the algorithm analyzes the electric field patterns produced by previous stimulations and then selects the next location that is most different from all others. This ensures that each new stimulation provides maximally new information.
To test this approach, we combined theoretical analysis, computer simulations, and experimental measurements in 10 healthy participants. We then compared FPS with the standard random sampling strategy.
Key results
Our findings show a clear advantage of the optimized approach:
Twice as efficient: FPS required only about half as many stimulation trials as random sampling to achieve the same mapping accuracy.
More robust: The FPS method performed more consistently across participants and was less sensitive to noise in the measurements.
Outlook
These results demonstrate that farthest point sampling can substantially improve the efficiency and reliability of TMS motor mapping. This opens the door to shorter experiments, reduced participant burden, and more practical applications of TMS in both research and clinical settings.
This work was carried out through an interdisciplinary collaboration across multiple departments at the University of Freiburg, in close cooperation with international partners from Tromsø, Norway. Zsolt Turi, Srilekha Marmavula, and Andreas Vlachos (Vlachos Lab, Department of Neuroanatomy, Institute of Anatomy and Cell Biology, University of Freiburg) David L. Schultheiss and Joschka Bödecker (Neurobotics Lab, Department of Computer Science, University of Freiburg) Peter Christoph Reinacher, Theo Demerath, Jakob Straehle (Department of Stereotactic and Functional Neurosurgery | Department of Neuroradiology | Department of Neurosurgery, Medical Center – University of Freiburg) Matthias Mittner (Institute for Psychology, UiT-The Arctic University of Norway Tromsø).
View the full article via PubMed or Imaging Neuroscience.
