Classification And Detection Of Skin Cancer Using Deep Learning Methods

The types of Skin Diseases are almost universal around the world because they are caused by genetic and environmental factors. Skin cancer has a very high mortality rate but is usually curable with minor surgical intervention [1]. Diagnosis of melanoma begins with clinical screening followed by dermatoscopy and histopathological examination. Melanoma skin cancer is highly curable if treated early buy detection is hard, the medical procedures are too long, and the doctor’s diagnosis cannot be 100% accurate. High speed cameras have 60-80% high accuracy stated by International Skin given Imaging Collaboration (ISIC). The aim of this project is to determine an accurate prediction of skin cancer and classify skin cancer as malignant or non-malignant. A quick and accurate diagnosis can be of great benefit to doctors and patients. In many areas of medicine, the performance of recent models based on deep learning approaches or even exceeds that of human experts. In this study, to help of dermatologists improve the efficiency of skin cancer analysis, we will use Convolutional Neural Networks (CNN) with Convnet model to complete the recognition of skin lesions images. This project aims to shorten the limitations by providing the predictive model using CNN to classify 9 types of skin cancer lesions from outlier lesion images. Proposal models are trained and evaluated using international standard dataset.