Title page for ETD etd-05052005-150740

Type of Document

Thesis

Author

Rajagopalan, Vidya

URN

etd-05052005-150740

Title

Neural Network based Prognosis System for Two-Dimensional Tumor-Like Growth

Degree

Master of Science

Department

Electrical and Computer Engineering, Department of

Advisory Committee

Advisor Name	Title
Simon Foo	Committee Chair
Anke Meyer-Baese	Committee Member
Namas Chandra	Committee Member
Uwe Meyer-Baese	Committee Member

Keywords

Wavelets
Neural Networks
Prediction

Date of Defense

2005-04-25

Availability

unrestricted

Abstract

In experimental and computational science and engineering, there are a host of practical problems where some tell-tale signs appear well before a critical event of interest surfaces, e.g., the occurrence of tumor is preceded by gradual change in cell density in the neighborhood. If these signs are detected then it is possible to prognosticate the severe conditions that are likely to occur. In this case the tumor can be predicted well before it actually appears and appropriate counter measures can be taken at an earlier stage with a higher degree of success. The effectiveness of such a system is gauged by how early the changes are detected and the accuracy of the predictions. The feasibility of using a combination of wavelets and neural networks as a prognostic tool is explored in this thesis.

There are two phases to implementing a prognosis system for tumor-like growth. The first phase involves characterizing the data so that the tell-tale signs are detected accurately. The next phase involves choosing an appropriate predictive tool to accurately model the growth of the tumor. If this system is built with some learning capability then it could emulate if not surpass the acumen of an expert. Wavelet analyses of incoming signals serve as a preprocessing tool while the neural network is used as the predictor. This wavelets-neural network prognosis tool is directly applied to detect the development of a tumor, well before an expert human eye can perceive the problem. Computer simulations are performed using real and simulated data sets, and conclusions are drawn from the results.

Files

Filename		Size	Approximate Download Time (Hours:Minutes:Seconds)
Filename		Size	28.8 Modem	56K Modem	ISDN (64 Kb)	ISDN (128 Kb)	Higher-speed Access
	thesispdf.pdf	891.48 Kb	00:04:07	00:02:07	00:01:51	00:00:55	00:00:04