Skin cancer, the most threatening skin disease, can spread to other parts of the body. It’s pervasive and aggressive damage can lead to death therefore most regular form of cancer. Over the last few years, there has been a rise in the reports of skin cancer in Asian continents. Early diagnosis of skin cancer in initial stages may augment the chance of survival. Regular skin checkups are recommended by dermatologist to identify the skin cancer in their initial stages. Hence, to assist this process, we proposed a mobile application that can detect the position of cancer and also classify into three categories such as Melanoma, Dermatofibroma, and Benign Keratosis lesions. In this paper, we proposed a convolutional neural network and implemented two models – Modified Inception model and Modified Google’s MobileNet with transfer learning. The evaluation of the proposed method is done using HAM10000 dataset which is a collection of multi-source dermatoscopic images of common pigmented skin lesions. The experimental results shows that modified inception model performs better than Google’s MobileNet. The objective of this paper is to develop a commercial mobile application to detect the chances of early cancer so that a proper treatment can be suggested to the patient.
Volume 11 | 07-Special Issue
Pages: 253-259