Blog
A Comprehensive Artificial Intelligence Framework for Dental Diagnosis and Charting
Authors: Tanjida Kabir, Chun-Teh Lee, Luyao Chen, Xiaoqian Jiang, and Shayan Shams
Summary
Kabir et al. proposed a comprehensive artificial intelligence framework for automatic tooth detection, numbering, and arrangement across panoramic and intraoral radiographs (periapical and bitewing).
The model employs U-Net and U-Net ResNet-34 architectures to segment teeth, bone regions, and the cementoenamel junction (CEJ) line, followed by a multi-scale matching process using either the patient’s panoramic image or a curated tooth repository built under the FDI numbering system.
The approach requ...
DOI: https://doi.org/10.1186/s12903-022-02514-6
Tooth numbering, Dental radiographs, Panoramic, Periapical, Bitewing, Deep Learning, Segmentation, Computer Vision, FDI System, Clinical Charting, AI in Dentistry, Diagnosis Framework
This work paves the way for fully automated, multimodal diagnostic pipelines integrated with electronic health records — offering faster, consistent, and explainable AI results for daily dental practice.
⚠️ Editorial note on dataset modality:
Although the framework integrates multiple radiographic modalities (panoramic, periapical, and bitewing), the datasets used were not multimodal in the strict sense. Panoramic radiographs came from a public repository (Abdi et al., 2015), while intraoral radiographs were collected from a private institutional database (UTHealth). The study did not specify whether the images belonged to the same patients. Therefore, the system represents a multimodal AI framework trained on unimodal datasets rather than a true paired multimodal dataset.