Dissertation/Thesis Abstract

Classification of Placental Chorionic Surface Vasculature Network Features Using Machine Learning Techniques
by Hambarsoomian, Hike, M.S., California State University, Long Beach, 2017, 50; 10603969
Abstract (Summary)

The placenta is an organ that connects the fetus to the uterine wall of the mother. Analyzing the Placental Chorionic Surface Vasculature Network (PCSVN) has found measurable anatomical indicators that appear to differentiate placentas associated with high and low risks for Autism Spectrum Disorders (ASD). Since vessels and nerves share many guiding mechanisms when they are created and with autism being a neurodevelopmental disease, the idea that the vasculature is different in the case of autism is plausible. With this thesis, we aim to improve understanding of the PCSVN factors which are visible already at birth or even earlier to identify children that are associated with placentas at risk for ASD. Using an embedded feature selection method called Elastic Net we were able to reduce the number of features by selecting the 16 most important of the original 66 PCSVN features. Using Principal Component Analysis, the dimension of data set was further reduced to five features (nodes, tortuosity, thickness, branching angle, and growth). We then use these five features to cluster and classify the placentas into the high and low risk cohorts. Because early diagnosis and treatment reduces the effects of ASD considerably, this thesis can be used to identify the high-risk cluster earlier than before, allowing children to begin treatment as soon as the placenta is classified.

Indexing (document details)
Advisor: Chang, Jen-Mei
Commitee: Ziemer, William, von Brecht, James
School: California State University, Long Beach
Department: Mathematics and Statistics
School Location: United States -- California
Source: MAI 56/05M(E), Masters Abstracts International
Subjects: Mathematics
Publication Number: 10603969
ISBN: 978-0-355-23018-5
Copyright © 2020 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy