Award Date

1-1-1998

Degree Type

Thesis

Degree Name

Master of Electrical Engineering (MEE)

Department

Electrical and Computer Engineering

First Committee Member

Lori Bruce

Number of Pages

84

Abstract

At present early detection is critical for the cure of breast cancer. Mammography is a breast screening technique which can detect breast cancer at the earliest possible stage. Mammographic lesions are typically classified into three shape classes, namely round, nodular and stellate. Presently this classification is done by experienced radiologists. In order to increase the speed and decrease the cost of diagnosis, automated recognition systems are being developed. This study analyses an automated classification procedure and its sensitivity to wavelet based image compression; In this study, the mammographic shape images are compressed using discrete wavelet compression and then classified using statistical classification methods. First, one dimensional compression is done on the radial distance measure and the shape features are extracted. Second, linear discriminant analysis is used to compute the weightings of the features. Third, a minimum distance Euclidean classifier and the leave-one-out test method is used for classification. Lastly, a two dimensional compression is performed on the images, and the above process of feature extraction and classification is repeated. The results are compared with those obtained with uncompressed mammographic images.

Keywords

Automated; Compression; Discrete; Effects; Mammographic; Recognition; Shape; Wavelet

Controlled Subject

Biomedical engineering; Electrical engineering; Diagnostic imaging

File Format

pdf

File Size

1976.32 KB

Degree Grantor

University of Nevada, Las Vegas

Permissions

If you are the rightful copyright holder of this dissertation or thesis and wish to have the full text removed from Digital Scholarship@UNLV, please submit a request to digitalscholarship@unlv.edu and include clear identification of the work, preferably with URL.

Identifier

https://doi.org/10.25669/q5jv-m9f1


Share

COinS