Detection of Eye Melanoma by Using Artificial Neural Network

Abstract

Melanoma of the eye is the most common type of cancer affecting the eye; although it's still quite rare. It occurs mainly as a secondary tumor from lung and breast cancers which are considered the most common cancers in the world. Even though early detection increases the chances of survival, problems occur due to difficulty in diagnosis. Artificial Neural Network (ANN) is a software program that mimics a biological neural network. It can be programmed to help in classifying and diagnosing the type of cancer. The objective of this research is to program ANN to classify and diagnose the type of eye melanoma. 40 pre-diagnosed samples (20 malignant and 20 benign) were used for this research. The images obtained were processed and enhanced using median filter and histogram techniques, and the region of interest which was determined by a physician was cropped and texture features were extracted from it using MATLAB. Finally using ANN the cancer was classified as either benign or malignant. ANN achieved an accuracy of 85 %, a sensitivity of 80 % and a specificity of 90 %. Furthermore the system was tested using a pre-diagnosed malignant image, which tested positive using ANN. ANN is effective in the classification of eye melanoma and has achieved high levels of accuracy. It’s recommended that physicians use ANN to enhance diagnosis, but a training course might be needed to help with the use of the program.