AI automatically detects and numbers teeth in periapical X-rays
A deep learning approach to automatic teeth detection and numbering based on object detection in dental periapical films
Authors: Hu Chen, Kailai Zhang, Peijun Lyu, Hong Li, Ludan Zhang, Ji Wu, Chin-Hui Lee
Summary:
A Faster R-CNN model was applied for automatic teeth detection and numbering in dental periapical radiographs. Using 1,250 annotated images, the network achieved over 90% precision and recall and a mean IOU of 0.91. Domain-specific post-processing (overlap filtering, template-based correction, and missing-tooth prediction) further improved numbering accuracy. Performance was comparable to a junior dentist’s level, highlighting AI’s role in supporting dental diagnostics.
Check the original publication:
Key Words: Deep learning, Faster R-CNN, Object detection, Tooth numbering, Periapical radiographs, TensorFlow, Dental AI, FDI system
Extracted Data
Year: 2019
**Modality: **Periapical Radiographs
Dataset: 1,250 images (800 train / 200 validation / 250 test)
Dataset Split: 800 / 200 / 250
Network Architecture: Faster R-CNN with Inception-ResNet-V2
Metrics: Precision: > 90%, Recall > 90%, Mean IOU = 0.91 ± 0.04
**AP – Professional Qty: **1
AP – Supervisor Presence: No information
AP – Experience Level: > 5 years
AP – Expertise Area: Oral Maxillo-facial Radiology
AP – Tool or System: No information
**ML Task: **Object detection and classification
Project Objective: Automate tooth detection and FDI numbering in periapical X-rays
Clinical Relevance
AI systems for automatic tooth numbering in periapical films can streamline dental diagnostics, minimize manual errors, and facilitate large-scale dataset creation. Faster R-CNN with dental domain post-processing provides a practical and efficient framework for clinical and research use.
Watch full video on YouTube
