Related Experiment Video
Updated: Mar 20, 2026
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
Developing and evaluating multimodal large language model for orthopantomography analysis to support clinical
Xinyu Liu1, Kuo Feng Hung2, Weihao Yu1
1Department of Electronic Engineering, The Chinese University of Hong Kong, Hong Kong SAR, China.
ToothXpert, a new AI system, enhances dental diagnosis using multimodal large language models. It provides interactive support for orthopantomography (OPG) analysis, improving accuracy and efficiency over existing AI and human dentists.
Area of Science:
- Artificial Intelligence in Dentistry
- Medical Imaging Analysis
- Multimodal Large Language Models
Background:
- Orthopantomography (OPG) is a crucial initial dental diagnostic tool.
- Current AI support for OPG lacks interactivity, unlike conversational AI models.
- There is a need for advanced AI systems that can provide dynamic and interactive dental diagnostic support.
Purpose of the Study:
- To introduce ToothXpert, an interactive dental diagnostic system utilizing a multimodal large language model (MLLM).
- To develop a comprehensive multimodal OPG dataset (MM-OPG) for training and evaluation.
- To assess ToothXpert's performance in dental diagnosis compared to existing AI models and human dentists.
Main Methods:
- Curated a dental concept alignment dataset and a large-scale multimodal OPG dataset (MM-OPG) with 131,065 question-answer pairs covering 11 conditions.
- Developed ToothXpert, an MLLM-based system enabling simultaneous visual and language responses for OPG analysis.
- Evaluated ToothXpert on internal (4,950 QA pairs) and external datasets, comparing its performance against other MLLMs and dentists.
Main Results:
- ToothXpert achieved a macro F1 score of 78.61% on the internal test dataset, significantly outperforming LLaVA v.1.5, LLaVA-Med v.1.5, Qwen-VL, and HuatuoGPT-Vision.
- Human assessment showed ToothXpert's averaged performance (3.54) surpassed compared MLLMs (1.46 and 1.38).
- On an external dataset, ToothXpert demonstrated higher F1 scores than junior dentists and required less diagnostic time.
Conclusions:
- ToothXpert represents a significant advancement in AI-driven dental diagnostics, offering interactive and accurate analysis of OPG images.
- The system's multimodal capabilities and extensive dataset contribute to its superior performance.
- ToothXpert has the potential to augment dental professionals' capabilities, improving diagnostic efficiency and accuracy.
More Related Videos
07:32Author Spotlight: 3D Movement Assessment of Maxillary Posterior Teeth in Clear Aligner Treatment
Published on: February 23, 2024
10:23Author Spotlight: Three-Dimensional Cephalometric Landmark Annotation Demonstration on Human Cone Beam Computed Tomography Scans
Published on: September 8, 2023