Derin öğrenmeyi kullanarak panoramik diş röntgenlerinden teşhis koymak

Radiology in dentistry is the first step in detecting past treatments, diseases, and problems that may arise in the future in the area of the mouth, teeth, and jaw. Thanks to the panoramic X-rays taken, the patient's current condition is considered as a whole, and the necessary treatments are planned because of the doctor's examinations. But because of the growing population, an insufficient number of doctors, insufficient hospitals, and economic difficulties, patients painfully wait a lot of rows first for an X-ray, and then for a doctor's check-up. In this thesis, the determination of the patient's past treatments from panoramic dental X-rays by the deep learning method is presented. It is expected that the study will accelerate doctors to find the patient's past diagnoses and be a preliminary preparation for finding diseases and problems that may arise in the future. As part of this study, previous studies that detected diseases from 2-dimensional or 3-dimensional X-rays conducted in the field of health were examined. It has been observed that the CNN algorithm is widely used as a deep learning method, and the YOLO algorithm is widely used for object detection, considering efficiency and speed factors. In some studies, techniques such as RPN for zone suggestion and SSD for object detection, Faster R-CNN have been used. In this study, the revised YOLO versions and the revised Faster R-CNN model were tested on a computer with the same capacity using the same data set. In this way, important factors such as success rate and speed were compared between these methods.