Refining Brain Stimulation Therapies: An Active Learning Approach to Personalization
Virtual: https://events.vtools.ieee.org/m/480226Brain stimulation shows significant potential in treating neurological disorders, but the challenge lies in personalizing these therapies effectively. Traditionally, identifying the optimal stimulation parameters, such as amplitude, frequency, and pulse width, requires extensive trial-and-error testing, which is both time-consuming and costly. To streamline this process, we developed an active learning framework that efficiently identifies the most effective relationships between stimulation parameters and brain responses, reducing the need for numerous experiments. We conducted three types of validation for our framework: in silico experiments using synthetic data from a Parkinson’s disease model, in silico tests with real data from a non-human primate model, and in vivo tests through real-time optogenetic stimulation in rats. In each scenario, our active learning models demonstrated superior performance over traditional random sampling methods, achieving significantly lower errors in predicting brain responses. This innovative approach enhances the efficiency and efficacy of research and clinical applications in brain stimulation, offering a more cost-effective pathway to developing personalized therapies for neurological disorders. Speaker(s): Dr Mohammed Sendi Virtual: https://events.vtools.ieee.org/m/480226