Cover Page    Full-Text Download    
Subscribe Now
Recommend the Paper
An Optimized, Economical and Painless Artificial Intelligence Technique to Diagnose Melanoma  

J.Abdul Jaleel, Sibi Salim, Aswin.R.B

Dept. of Electrical & Electronics Engineering, TKM College of Engineering,

Kollam-695005, Kerala, India

 drjaleel56@gmail.com, sibi_salim@rediffmail.com,aswinrb@gmail.com

 
Abstract .Skin cancer is a deadly condition occurring in the skin. It is a gradually evolving condition which starts in the melanocyctes in skin. So it is also called as Melanoma. First it occurs in a small region and later spreads to other parts of the body through the lymphatic system. If the skin cancer is detected at the early stages, it can be cured. So an early detection system is inevitable in skin cancer diagnosis. Melanoma can be of Benign or Malignant. Malignant melanoma is the dangerous condition, while benign is not. At initial stages, both of them resembles in appearance. So a classification of benign and malignant melanoma is difficult. Only an expert Dermatologist can make correct classification. Conventional diagnosing procedures include preliminary diagnosis by direct observation by doctors and Biopsy method for confirmation. Biopsy method is a painful and time consuming one. So an efficient classification system using Artificial Intelligence (AI) and Image Processing Techniques (IPT)is proposed. Dermoscopic images are given as input to the system. Images contain noises and hairs. The noises are removed using image processing techniques. After that, region of interest or suspicious region of skin is separated from normal skin using Segmentation. Segmentation method used here is Color Threshold Segmentation. Two feature extraction techniques used- Gray Level Co-occurrence Matrix (GLCM) method and Red, Green, Blue (RGB) color features. These features are gives as the input to Artificial Neural Network Classifier. It classifies the given data set into Cancerous and Non-cancerous.
 
Keywords : Skin cancer, Segmentation, Gray Level Co-occurrenc
 URL: http://dx.doi.org/10.7321/jscse.v3.n3.86  
 
 

Subscribe Now

Email :    
Subscribe to receive free TOC's JSCSE by email
Subscribe

Recommend To Friend

Email :     People