Human Skin Diseases Detection and Classification using CNN
Shamim Ahmed, MD. Abiduzzaman, Mahedi Hasan Rajib, Rafat Rahaman, Salam Hussain, Hafsha Binta Rashid, Md. Anwar Hussen Wadud, T. M. Amir-Ul-Haque Bhuiya
Abstract
Recently skin disease has turned into a threat to normal human life when it becomes cancer. If there do any negligible steps during treatment it will become severe and also may cause the reason of human death. So, necessary steps in an early stage of the diagnosis are mandatory to avoid the worst scenario like human death, especially when the disease is Melanoma skin cancer. Because the human death rate of Melanoma skin cancer has become high during this recent era. To detect skin cancer disease there is some common treatment strategy which is done by expert healthcare professional like doctor or dermatologist, as well as these procedures are also very lengthy and time-consuming. Therefore, recently some automated computer vision or image processing with artificial neural network-based methods are introduced but during the implementation level, these methods are highly cost-effective thoughthey provide better accuracy during prediction. In this paper, we proposed a deep learning-based CNN method for classifying skin diseases. To implement our proposed model, we gathered approximately 2102 image datasets from a massive repository of diverse source images, particularly dermatoscopic images of pigmented lesions. Our experiment results show an 89.45 percent accuracy in detecting skin cancer disease. We hope our proposed method will help to diagnose this skin cancer disease more efficient way than other methods that have already been used.