Optimizing electrophysiological techniques for localizing deep brain structures.

Deep brain stimulation (DBS) is an established treatment for Parkinson's disease and dystonia, with exciting possibilities for applications in other movement disorders, psychosurgeries, and brain pathologies. Accurate targeting of the implant site is the most important predictor of a positive DBS outcome, and is also the most time-intensive portion of the procedure. The current standard for localizing the subthalamic nucleus (STN), the most prevalent DBS target for treating Parkinson's disease, is a multi-step procedure including an electrophysiological confirmation phase during which microelectrode recorded signals in the vicinity of the target are observed to identify the neural signatures of the brain structures. This dissertation utilizes various engineering tools and quantitative analyses to improve the electrophysiological identification portion of the STN target localization process and ultimately increase the speed and accuracy of DBS procedures. We demonstrate that using independent component analysis to pre-process neural data recorded with close-spaced multichannel electrode arrays quantifiably improved the signal-to-noise relationship on neural recordings taken within the thalamus/subthalamus of a rat. This improvement resulted in more accurate spike detection. We identified a set of quantitative localization parameters to demarcate structural boundaries and define the signature of the subthalamic nucleus in an objective way. One such parameter was a wavelets-based measure of the magnitude of the background non-spike activity on neural data records, and this parameter proved successful at locating the boundaries of the STN during offline analysis of recordings taken during the electrophysiological mapping portion of 14 Parkinson's disease patients undergoing DBS surgery at University of Michigan hospital. An index comprised of contributions from five different parameters defined the STN such that a sufficiently positive score on this index correctly identified individual recordings as having originated in the STN in 90 of the 91 instances that the score was reached. The final experiment presented here demonstrated the feasibility of combining multichannel electrode recordings, ICA processing, and localization parameters to identify structural boundaries in a rat. Using an animal model allowed histological verification of the structure localization. Also, ICA processing improved the quality of the data, which led to more accurate calculation of the spike train based parameter tested (firing rate).PhDApplied SciencesBiomedical engineeringUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/126563/2/3253407.pd