Loading...
Computer based analysis of mammogram images to treat breast cancer
Mitra, Parthib
Mitra, Parthib
Citations
Altmetric:
Files
Loading...
t16081_Mitra.pdf
Adobe PDF, 4.7 MB
Authors
Other Names
Location
Time Period
Advisors
Original Date
Digitization Date
Issue Date
2016-12
Type
Thesis
Genre
Keywords
Subjects (LCSH)
Electronic thesis
Electronic dissertations
Electronic dissertations
Citation
Abstract
Breast cancer is a curse to complete human race. It grows inside breast cells silently, and soon it spreads to the whole body; as a result, the patient dies very soon. According to recently published reports, about 12.5% women are suspected of developing invasive breast cancer during their lifetime in the United States. Mammography, an imaging technique, is used in most breast cancer detection methods. A typical mammogram has poor contrast; as a result, doctors often overlook microcalcifications while using mammograms. In this study, we introduce a computer-based image analysis methodology to examine mammograms. Our proposed methodology selects doubtful areas on images, extracts hidden attributes of the selected areas, assesses the extracted values, and helps classify the images as benign or malignant. A simulation platform is developed using popular MATLAB tool to evaluate the proposed methodology. Open source mammogram images from Digital Database for Screening Mammogram (DDSM) and Mammographic Image Analysis Society (MIAS) are used to run the simulation programs. Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM) algorithms are used to analyze the extracted feature values. The proposed methodology has potential to accurately examine the suspicious regions in mammograms, because the suspicious regions are converted into equivalent digital values. Experimental results suggest that feature extraction is a promising way to analyze mammogram images to treat breast cancer. LDA and SVM methods provide 100% accuracy while classifying the images. It is observed that the geometrical features (such as radius of the selected area) have less or zero impact in detecting cancer. However, the textural features (such as standard deviation of the extracted values) have significant impact in classifying the images as malignant or benign. This work can be extended to study other types of cancer such as lung and ovarian cancer.
Table of Contents
Description
Thesis (M.S.)--Wichita State University, College of Engineering, Dept. of Electrical Engineering and Computer Science
Publisher
Wichita State University
Journal
Book Title
Series
Digital Collection
Finding Aid URL
Use and Reproduction
Copyright 2016 by Parthib Mitra