Five Layered Mask-RCNN based Dental Disease Detection from Children Panoramic Radiographs

Dentists typically use their naked senses to diagnose oral diseases, however this method presents numerous potential for misinterpretation and incorrect evaluations. The periodontal illness identification assistance platform based on advanced learning algorithm is created for increasing the precision and performance of the diagnosis of dental disorders. The technology could be extending its service that can autonomously recognize the dental images gathered by the recording device that aids the dentist with the diagnosis. This paper suggests Five Layered Mask-RCNN (FLM-RCNN) to diagnose the children dental disease from the Panoramic Radiograph images. The proposed model uses Children's Dental Panoramic Radiographs Dataset from Kaggle that consists of 1000 dental images. The FLM-RCNN model initiates with exploration of image distribution. The dental images are fitted with Mask RCNN model that performs the segmentation, selecting the ROI region and performs the bounding box regression for prediction of children dental disease. The existing Mask-RCNN is fitted with dental images by varying the number of convolutional layers. The dental images are applied to FLM-RCNN which is designed with five convolutional layers and the same images are fitted with Mask-RCNN with 2, 3, 4, 6 and 7 convolutional layers. The experimental results explores that the recommended FLM-RCNN with five convolutional layer exhibits highest Accuracy of 98.42 when examined against other convolutional layers.