Date of Award

Fall 2014

Document Type

Thesis

Degree Name

Master of Science in Electrical Engineering - (M.S.)

Department

Electrical and Computer Engineering

First Advisor

Atam P. Dhawan

Second Advisor

Yun Q. Shi

Third Advisor

Edwin Hou

Abstract

Effective screening to detect the skin cancer accurately in the early stage is essential for reducing the mortality of skin cancer. Surface features, such as texture and pigmentation area from the surface, epi-illumination images of the skin lesions have been well correlated to detect skin cancer. An increase in the lesion's subsurface blood volume has been correlated to early diagnosis of malignant melanoma. A method for estimating the optimal features is obtained. The optimal features help in accurately classify the skin lesion in various grades. To make the process faster these optimal features are clustered. The optimal clusters are obtained by genetic algorithm. The optimal cluster centers act as input to the SVM classifier and the kernel parameters are obtained. Finally, parameters of the kernel function are optimized by genetic algorithm, which help in classifying the skin lesions into various grades leading to early diagnosis of skin cancer.

Share

COinS