A comparative analysis of melanoma detection methods based on computer aided diagnose system

Abstract

Skin cancer is a very common type of cancers. It is increasing day by day. Among all skin cancers melanoma falls in most dangerous category. Melanoma has caused highest mortality rate. It can be cured if it is diagnosed earlier. In 2020 there will be 6850 death due to melanoma. Computational methods are proved to be a backbone for early detection of this deadly disease. It is very hard to differentiate between melanoma and non melanoma lesion due to low contrast and high degree of similarity in visualisation. Computer aided diagnose (CAD), a non invasive technique is seen to improve early and faster identification of melanoma. CAD has four steps: Noiseless Pre processing, trustworthy lesion segmentation, relevant feature extraction and an accurate classifier. This paper presents main and current computational methods for skin cancer recognition with dermoscopic images. This paper also presents statistics and results by comparing the performance of classifiers as well as pre processing, segmentation and feature extraction methods.