Deep learning based solutions to biomedical image analysis problems of the upper aerodigestive tract
No Thumbnail Available
Authors
Meeting name
Sponsors
Date
Journal Title
Format
Thesis
Subject
Abstract
[ACCESS RESTRICTED TO THE UNIVERSITY OF MISSOURI--COLUMBIA AT REQUEST OF AUTHOR.] Clinical and scientific studies produce an abundance of image and video data. Clinical practice often relies on subjective, qualitative interpretation of image and video data, whereas scientific studies often involve extremely labor intensive manual analysis methods creating a bottleneck in scientific discovery. Efficient, objective, quantitative computational image and video analysis tools are needed to facilitate disease and treatment monitoring, early diagnosis, and scientific discovery. In this work, we explore artificial intelligence and machine learning guided computer vision frameworks and solutions for objective and quantitative analysis of biomedical image and video data. Of particular interest to this study is the analysis of upper aerodigestive tract motion behavior that is responsible for the life-sustaining functions of breathing and intake of food. This is the first study of its kind that generates quantitative outcome measures from endoscopic videos of vocal folds using flexible endoscopy. This study also reduces the bottleneck in analysis of videofluoroscopic swallow study (VFSS) videos with rodent models of swallowing impairment. The proposed solutions involve semi-automated, verifiable visual tracking solutions for immediate clinical use and for training data generation, novel deep learning network cascades with location priors for classification, robust segmentation schemes combining region-based and contour-based clues (glottal region segmentation), and ensemble of deep networks for laryngeal adductor reflex (LAR) detection. Novel multi-task deep learning solutions were developed to address the various needs in upper aerodigestive tract function analysis.
Table of Contents
DOI
PubMed ID
Degree
Ph. D.
Thesis Department
Rights
Access to files is limited to the University of Missouri--Columbia.
License
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 License. Copyright held by author.