Background: Microelectrode recordings (MERs) of the neural activities are a useful tool for subthalamic nucleus localization (STN) in the process of deep brain stimulation (DBS) surgery for neurological and neuropsychiatric disorders, like Parkinson's disease. Currently, the localization of STN relies on manual demarcation performed by a neurophysiologist. It is also di cult to give an exact detection of the STN borders, especially the one between the STN and the substantia nigra pars reticulata (SNr). As a result, investigators are looking for a multi-feature machine learning method for accurate automated STN localization. We used K-means clustering and a novel set of features that are based on a combination of spike-based and spike independent metrics to address these shortcomings. Methods: We extracted 18 computational features from 31 bandpass ltered (0.5 kHz - 8 kHz) MERs of 20 patients who underwent DBS implantation surgery. 12 of these 18 features were spike-based and the other 6 were spike independent. Square root transformation was performed on those positively skewed features. All features were standardized and normalized to scale [0, 1]. We additionally integrated a depth vector into the original feature matrix. The newly formed feature matrix was fed into the k-means clustering algorithm. K was chosen to be 4 because the trajectories, under most circumstances, went through four consecutive structures in the brain. The STN demarcations were predicted by a combination of k-means clustering and feature activity maps. The results were later compared to the decisions of a human expert. Results: Two of the four newly designed features, i.e. mean inter-spike interval and bursting rate, showed good discriminatory power between STN-SNr region and all the other regions. The modi ed k-means clustering with k = 4 identi ed the STN entry and STN exit with an error of 0:0443 0:2190 and 0:2555+-0:4005, respectively (mean +- standard deviation), and achieved detection sensitivity (error < 0.5 mm) of 93:5% and 80:6%, respectively. Conclusion: The combination of k-means clustering and a novel feature activity map provided a simple and robust method for STN localization. Further integration of the depth information into the original feature matrix has substantial effect in reducing the noise existed in the clustering results given by k-means clustering alone. Since our unsupervised learning approach does not require manual demarcation on its output, the degree of automation has improved. This is expected to lead into a decreased time spent inside the operation room and hopefully improve the clinical outcomes.
Subject (authority = RUETD)
Topic
Computer Science
RelatedItem (type = host)
TitleInfo
Title
Rutgers University Electronic Theses and Dissertations
Identifier (type = RULIB)
ETD
Identifier
ETD_8518
PhysicalDescription
Form (authority = gmd)
electronic resource
InternetMediaType
application/pdf
InternetMediaType
text/xml
Extent
1 online resource (ix, 31 p. : ill.)
Note (type = degree)
M.S.
Note (type = bibliography)
Includes bibliographical references
Note (type = statement of responsibility)
by Fan Yang
RelatedItem (type = host)
TitleInfo
Title
School of Graduate Studies Electronic Theses and Dissertations
Identifier (type = local)
rucore10001600001
Location
PhysicalLocation (authority = marcorg); (displayLabel = Rutgers, The State University of New Jersey)
I hereby grant to the Rutgers University Libraries and to my school the non-exclusive right to archive, reproduce and distribute my thesis or dissertation, in whole or in part, and/or my abstract, in whole or in part, in and from an electronic format, subject to the release date subsequently stipulated in this submittal form and approved by my school. I represent and stipulate that the thesis or dissertation and its abstract are my original work, that they do not infringe or violate any rights of others, and that I make these grants as the sole owner of the rights to my thesis or dissertation and its abstract. I represent that I have obtained written permissions, when necessary, from the owner(s) of each third party copyrighted matter to be included in my thesis or dissertation and will supply copies of such upon request by my school. I acknowledge that RU ETD and my school will not distribute my thesis or dissertation or its abstract if, in their reasonable judgment, they believe all such rights have not been secured. I acknowledge that I retain ownership rights to the copyright of my work. I also retain the right to use all or part of this thesis or dissertation in future works, such as articles or books.