Dissertation/Thesis Abstract

A filter bank approach to automate vessel extraction with applications
by Huynh, Nen, M.S., California State University, Long Beach, 2013, 123; 1523090
Abstract (Summary)

Methods to extract vessel networks in medical images have been in high demand for its applications to health risk predictions. For example, vessel enhancement of retinal images has shown promises in diagnosing diabetes. Within the existing literature, multiscale vessel enhancement stands out as one of the best for its accuracy, speed, and simplicity. But like many vessel extraction techniques, the efficacy of the method is greatly hindered in the presence of noise, lighting variations, and decreased resolution. This deficiency is presents itself in retinal images and are particularly pronounced in digital photographs of human placenta.

Retinal images have a been popular data set of testing vessel extraction methods because of its simplicity in anatomical structure yet high hopes in diagnosing conditions such as diabetic retinopathy and glaucoma. Thus, the thesis will focus on the application of vessel extraction methods on retinal images. Specifically, we focus on the DRIVE and STARE database.

Also, recent placental pathology evidence has contributed to current understanding of causes of low birth weight and preterm birth, each has been linked to increased risk of later neurodevelopmental disorders. Among various factors that cause such disorders, the vessel network on the placenta has been hypothesized to offer the most clue in bridging that connection. Herein lies the most essential step of the blood vessel extraction, which has only been done manually through a laborious process.

Motivated by its ability to handle curvilinear structures, we propose the use of directional filter banks to further enhance the results obtained from the multiscale method. Validating experiments will be performed on a private database that is made available by the Placental Analytics, LLC.

It will be shown that for retinal images, the directional filter bank approach significantly improves the performance over the well-known multiscale vessel enhancement method. However, the directional filter bank approach are comparable to multiscale vessel enhancement on placentas.

Indexing (document details)
Advisor: Chang, Jen-Mei
Commitee: Kim, Eun Heui, Ziemer, William
School: California State University, Long Beach
Department: Mathematics and Statistics
School Location: United States -- California
Source: MAI 52/01M(E), Masters Abstracts International
Source Type: DISSERTATION
Subjects: Applied Mathematics, Mathematics, Health sciences
Keywords: Directional filter bank, Image processing, Placenta, Retina, Vessel enhancement
Publication Number: 1523090
ISBN: 9781303202940
Copyright © 2019 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest