Machine Learning for Programming Deep Brain Stimulation Arrays
Improved Neuromodulation Approaches, 2014
Study Rationale: † † † † † † † † ††
Deep brain stimulation (DBS) implants for treating Parkinsonís disease currently consist of four round electrodes stacked along a thin cylindrical carrier lead. When these DBS leads are implanted with millimeter-scale deviations from the intended brain target, stimulating current can spread beyond the target borders and induce adverse side effects. Recent development of a DBS lead with segmented electrodes holds significant promise. However, the process of optimizing the stimulation parameter settings for such DBS arrays will create significant clinical challenges as we seek to provide the best therapy possible for each patient.
We hypothesize that stimulation through round electrodes will cause radially uniform changes in neuronal firing rate and pattern around the DBS lead, whereas stimulation through columnar electrodes will result in asymmetric changes in neuronal activity around the DBS lead.
This project will: (1) develop structural machine learning and optimization routines based upon high-field imaging and computational neuron models for programming DBS arrays with radially segmented electrodes, and (2) apply and validate these algorithms in a pre-clinical model of Parkinsonís disease using quantitative motor testing and neuronal spike recordings around the implanted DBS arrays. In this study, the DBS arrays will be implanted through the subthalamic nucleus and globus pallidus, but along trajectories that would otherwise be considered suboptimal because of low-threshold motor side effects.
Impact on Diagnosis/Treatment of Parkinsonís Disease: † † † † † ††
The development and integration of these machine learning algorithms coupled with radially segmented DBS arrays will provide a transformative technology to improve the efficacy of DBS therapies for individuals with Parkinsonís disease.
Next Steps for Development:
This project will provide a translational pathway towards investigational use and then clinical trials of these novel programming algorithms for subject-specific optimization of stimulation settings through high-density DBS arrays for treating Parkinsonís disease.
Assistant Professor of Biomedical Engineering at University of Minnesota
Location: Minneapolis, Minnesota, United States