Detection and Classification of Dental Pathologies using Faster-RCNN in Orthopantomogram Radiography Image
Anuradha Laishram, Khelchandra Thongam
Abstract
An attempt has been made to device a robust method to detect and classify different oral/dental pathologies using Orthopantomogram (OPD) images based on Faster- Regional Convolutional Neural Network (Faster-RCNN). The system will provide a novel method for detection and classification of types of teeth (viz., incisors, molar, premolar, canine teeth) and also some underlying oral anomalies such as fixed partial denture and impacted teeth. Initially, various image pre-processing techniques are performed. For each input signal containing teeth of a person, detection is done using the concept of Anchors and Intersection over Union. The main advantage of our algorithm is the detection using bounding boxes which replace the manual separation of each individual tooth from the set of teeth of the signal. The proposed algorithm is implemented for training and testing and also compared with the ground truth and our model gives an accuracy of more than 90% for detection and 99% for classification.