An Efficient Skin Cancer Prognosis Strategy Using Deep Learning Techniques

Abstract

To start the right treatment, identification of an indefinite skin lesion is necessary. Only highly trained dermatologists who can treat them with an early diagnosis and diagnose melanoma skin lesions. The classification of skin for melanoma Dermoscopic images is 70 percent. Due to the limited supply of expertise, systems that sort dermal growth as the autoimmune or metastatic tumor can serve as an early screening tool. This study provides a model of the Convolutional neural network trained for skin lesion images, from previously acquired features of the Highway Convolutional neural network (CNN). It does not require advanced preprocessing. In addition, the model not require much computing power to train. The Convolutional neural network (CNN) method achieves training accuracy of 50%, and 70% of the test data have classification accuracy, low, moderate and high accuracy of the estimated damage.