Related Experiment Video
Updated: May 2, 2026
03:14
Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
Published on: December 6, 2024
1.3K
A benchmark multimodal oro-dental dataset for large vision-language models.
Haoxin Lv1, Ijazul Haq2,3, Jin Du4
1Department of Oral Implantology, Suzhou Doctor Dental Clinic Co. Ltd, Suzhou, 215000, China.
Scientific Data
|April 30, 2026
Summary
A new multimodal dataset of 8775 dental checkups enhances artificial intelligence in oral healthcare. Fine-tuned AI models show significant improvements in diagnosing oro-dental anomalies and generating reports.
Area of Science:
- Artificial Intelligence
- Oral Healthcare
- Medical Informatics
Background:
- Advancements in AI for oral healthcare require large, multimodal datasets.
- Existing datasets may not fully capture the complexity of clinical dental practice.
Purpose of the Study:
- To introduce a comprehensive multimodal dataset for AI in oro-dental healthcare.
- To evaluate the utility of this dataset by fine-tuning large vision-language models.
Main Methods:
- Collected 8775 dental checkups from 4800 patients (2018-2025).
- Dataset includes intraoral photographs, radiographs, and textual clinical records.
- Fine-tuned Qwen-VL 3B and 7B models for anomaly classification and report generation.
Main Results:
- Fine-tuned models demonstrated substantial performance gains over base models and GPT-4o.
- Validated the dataset's effectiveness for AI-driven oro-dental diagnostics.
- Achieved high accuracy in classifying oro-dental anomalies and generating diagnostic reports.
Conclusions:
- The presented multimodal dataset is a valuable resource for AI in dentistry.
- The dataset effectively enhances the performance of AI models in oral healthcare applications.
- Public availability of the dataset will foster future research in AI dentistry.
