Derin Öğrenme ile Deri Rahatsızlıklarının Erken Teşhisi için Bir Sistem Tasarımı.

Today, an increase has been observed in the number of deaths due to skin diseases. People who die from skin cancer, which is one of the diseases with a high death rate worldwide, constitute 30% of the total people who die from cancer. In addition, skin diseases are the harbingers of diseases that may occur in the body in the future. On the contrary, other studies have shown that early detection of skin cancer greatly saves the life of the patient. As a solution proposal for the diagnosis of skin diseases, a device that can be used to perform skin scanning has been proposed in the study. The device proposed in this study has been designed in a structure similar to the video dermoscopy device and works with the images taken instantly from the patient. These images have been given to the deep neural network and the outputs are both seen over the interface and presented as a report. A moving mechanism have been designed for the device to receive data from the human body. This moving mechanism has been scans the human body 360 degrees and detects the nevus. For the deep learning software of the study, the ISIC2018 dataset has been arranged and appropriate data have been selected and data augmented. A total of 1400 images have been trained with the ResNet-50 pre-trained model. As a result of this training, an accuracy of 96% was obtained. [ABSTRACT FROM AUTHOR]