Machine learning in the field of dentistry

Abstract

Due to the increasing availability of dental data, deep learning has been adopted to automate clinical tasks. Object detection using deep learning techniques has gained popularity in the field of dentistry due to the growing demand for automated diagnostic imaging. This study aimed to detect and number teeth and dental conditions such as root canal and implant on dental x-rays using a convolutional neural network, which provides fast and accurate results. 700 dental panoramic images were used in this study. Each image and tooth were annotated and categorized manually. Faster R-CNN with backbone network ResNet 101 was selected as it has the best performance at the COCO Object detection contest and is considered a state-of-the-art object detection model. The model was able to number and detect 34 classes (32 teeth, Root Canal, and Implant). It performed well, providing accurate detections with detection scores of more than 90% on test images that are comparable to a dental expert. A Graphical User Interface (GUI) was developed using the python library pyqt5 to allow users to perform analysis with various options using the model.