Skin Cancer Diagnosis Using High-Performance Deep Learning Architectures

Skin cancer is one of the dangerous diseases which causes human death. It is originated by the abnormal growth of melanocytic cells. The malignant Skin cancer is also referred by another name as melanoma. The key source for the computer aided diagnosis is Dermoscopy image which acquires the skin cancer. Fortunately, a high percentage of skin cancers can be cured if they are found early. Automatic early detection approaches are being used as quick and practical skin cancer screening due to the expense and morbidity of this biopsy procedure. The existing methods of melanoma detection meet-up the demerits such as less accuracy, false segmentation and more time consumption. To overcome these issues, the proposed High performance Deep Learning Architecture method for classifying the skin cancer. It involves three types of techniques are used for comparison study are MobileNet, ResNeXt101, and Multi-Model Ensemble. The performance metrics used for the comparison are accuracy, precision, and recall. The comparison of the accuracy value for MobileNet, ResNeXt101 and Multi-model ensemble techniques are 97.2%, 84.2% and 86% respectively.