Automated tooth number recognition from smartphone intraoral photographs using Mask R-CNN: internal and external validation study

Tooth numbering plays a significant role in dental care, providing accurate record keeping, diagnosis, and treatment of dental issues. The aim of this study was to evaluate the Mask Region-based Convolutional Neural Network (Mask R-CNN) model for tooth number detection, classification, and segmentation from smartphone photographs. This retrospective study utilized the data collected in Finland during 2022. A total of 448 individuals aged 13–78 years participated in this study. Participants uploaded dental photographs taken by participants’ own or research group’s smartphone. The dataset consisted of upper and lower occlusal, right and left lateral and frontal views. Teeth were labelled using polygon annotation according to the FDI notation system. The final number of images and tooth polygons were 1,272 and 14,736 respectively. The Mask R-CNN model was used for tooth number recognition. Performance of trained models was measured based on the tooth object detection for 28 tooth classes using PASCAL performance metric’s open toolset. The mean average precision (mAP) was calculated for each class with an intersection over union (IoU) threshold of 0.5, as well as precision, sensitivity and F1-score. External validation was conducted using open dataset from India, consisting of 39 preliminary selected images. The model demonstrated good performance with an overall mAP of 80.0%, with a standard deviation of 1.0% on 10-fold training stability validation, which indicates pipeline robustness for the internal dataset domain. The highest individual tooth detection performance in the internal test subset was recorded for the maxillary left canine, tooth number 23, with all evaluation metrics achieving 100%. The external validation further confirmed the model’s strong performance, yielding a mAP of 84.1%. The highest individual tooth detection performance in the external test subset was recorded for the mandibular left second incisor, tooth number 32, with all evaluation metrics achieving 100%. The Mask R-CNN model developed in this study showed good mAP for tooth localization and number recognition within smartphone photographs in internal and external testing supporting the potential generalizability of the model. The model developed can accurately localize and classify teeth from photographs taken by smartphone cameras possibly enabling remote digital dental screening.