Masters Theses

Abstract

“Cervical cancer is one of the most deadly cancers faced by women. It is the second leading cause of cancer death in women aged 20 to 39 years. In order to detect cancer at early stages, pathologists analyze the epithelium region from the cervical histology images. These histology images have a pre-cervical cancer condition called cervical intraepithelial neoplasia (CIN) determined by pathologists. This study deals with automating the process of epithelium detection and epithelium CIN classification in digitized histology images. For epithelium detection, the objective is to detect epithelium regions in microscopy images from non-epithelium regions and background. convolutional neural networks, both shallow and deep networks are used for epithelium detection. The highest epithelium detection accuracy of 98.84% is obtained using transfer learning on VGG-19 architecture, pre-trained on the ImageNet dataset. For CIN classification, the epithelium region is divided into 5 segments along the medial axis and patches from each segment were used for training the deep learning model. Vertical segment level classification probabilities from deep learning model are obtained and further classified using SVM, LDA, MLP, logistic and RF classifiers. The highest image level accuracy obtained is 77.27% for MLP classifier using voting”--Abstract, page iii.

Advisor(s)

Stanley, R. Joe

Committee Member(s)

Stoecker, William V.
Kosbar, Kurt Louis

Department(s)

Electrical and Computer Engineering

Degree Name

M.S. in Computer Engineering

Comments

The National Library of Medicine (NLM) supported this research.

Publisher

Missouri University of Science and Technology

Publication Date

Spring 2018

Pagination

ix, 42 pages

Note about bibliography

Includes bibliographic references (pages 39-41).

Rights

© 2018 Sri Venkata Ravitej Addanki, All rights reserved.

Document Type

Thesis - Open Access

File Type

text

Language

English

Thesis Number

T 11992

Electronic OCLC #

1313117335

Download

Share

 
COinS