An algorithm suitable for online identification of subthalamic nucleus from microelectrode recording data

This dissertation presents a new algorithm for online identification of the subthalamic nucleus (STN) using microelectrode recording (MER) during deep brain stimulation (DBS). DBS is a common surgical technique used to suppress Parkinson’s disease (PD) symptoms especially among late stage patients. Although magnetic resonance imaging (MRI) is capable of identifying the approximate location of STN, MER is still commonly used as a complementary measure due to its higher accuracy and easy intraoperative manipulation. Final decision of the electrode placement is still made by the neurosurgeon based on the response from patients when stimuli are given at different spots, in order to gain the best tremor control ability from DBS. However, algorithms that utilize the MER data from the tip of DBS electrode also provide valuable help for the neurosurgeon to make a more informed decision. Especially in situations where the response from patients is less distinctive across STN boundary, those algorithms can be critical for the success of DBS surgery. In this research we aim to develop an algorithm that is more robust and accurate. To accomplish this, we will focus on several different metrics based on the same recording. Various methods are used to evaluate and select these features. Then, a machine learning mechanism is used to cluster the data using fuzzy c-means clustering and identify STN based on step detection. The algorithm has very good sensitivity and specificity for the task it faces. It is also efficient and suitable for online application.Ph. D.Includes bibliographical referencesby Yuwen Zhan