Diagnosing melanoma cancer using deep learning

Skin cancer is common and a serious disease, which can lead to death if not treated in time. Melanoma is the rarest and most dangerous type of skin cancer. It causes the most deaths. As in all diseases, early and correct detection of skin cancer is very important. It is expected that melanoma cancer, which is aimed to be detected using deep learning in this project, will be of great convenience for the patient and the doctor. The main purpose of the project is to help the early diagnosis of patients with skin cancer and to save time for the health system, thus enabling the initiation of fast and effective treatment methods. Thanks to these developments in the field of health of artificial intelligence, physician and patient comfort increases. In this study, the data from the ISIC Archive were trained with a deep learning algorithm. Using the deep learning method, it was determined by the software that the 'nevus' in the body was malignant or benign over the appropriate camera apparatus. Thus, early diagnosis is provided, and patients will be able to consult their doctors. The developed algorithm has achieved a good result by detecting moles with nearly 100% success and the accuracy rate was 88% in the SSD MobileNet V2 method and 89% in the SSD ResNet101 method. With these conclusions, it is seen that the study will shed light on future studies and diagnosis systems based on mobile health. Key Words: Deep Learning, Melanoma Cancer, Artificial Intelligence, Early Diagnosis