Development of a deep learning based disease prediction system on clinical images of body moles

Scientifically, melanomas are less common than other skin cancers. However, approximately 75% of skin cancer-related deaths are due to melanomas. Cancer cells Show a behavior that grows very quickly metastasis systemically As with all types of cancer, early diagnosis is of great importance for the treatment of skin cancer. Therefore, it is extremely important to recognize and accurately evaluate the signs of cancer occurring on the skin early. Therefore, when a color change occurring on the skin or a change in the body me is noticed, it should be followed in terms of color and shape, blisters in skin tones or pink tones should also be considered. Dermatologists often recommend regular skin examination to identify and eliminate Melanoma in its early stages. In this study, skin cancer detection was made by examining the patterns on the colored lesion images on the patient's body with deep learning techniques. While the importance of early diagnosis in medicine has been emphasized many times, our study aims to diagnose early and move to the treatment stage if there is a risk of cancer. In our study, convolutional neural networks were successfully trained with an open source database with pathological validations with dermoscopic images. With this method, the moles in the body are created and the point areas are determined. In dermoscopic image devices; color images are recorded by creating a map of the moles in the body. Thus, the chance to compare with the image to be obtained in the next check is provided. In this multiclassification application, a high degree of success was achieved.