Deep Transfer-Based Skin Carcinoma Detection

Abstract

With the emergence of several pollutants, cosmetics, and chemicals into our day-to-day lives, skin cancer is becoming a common disease. Machine learning and image processing is used for identification of type of skin cancer. Several algorithms have been proposed to detect skin cancer, but most of the inputs are fed manually. Manual testing for skin cancer is difficult and strong similarities between different skin types can lead to false detection of lesions classes. To overcome this problem, we propose an algorithm which requires minimal intervention of doctors when provided with an input affected skin image. Images of the affected area are captured with the help of derma scope and fed to the model. The model initially checks whether the image contains a skin tissue that has cancer or not and if it has cancer then classify into which class of cancer it is, whether it is melanoma, nevus, or seborrheic keratosis. Skin cancer segmentation, along with the same evaluation criteria and the results, also showed that the individual accuracy of each class of skin cancer is computed in efficient manner. The models iteratively learn from its past experience and make the model more enhancing. By 2021, there will be 6.3 billion smartphone subscriptions, which might enable low-cost universal access to skin cancer diagnostic services.