Multimodal Language Model for Jaw Osteonecrosis Diagnosis and Treatment

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oral and Maxillofacial Surgery

Background:

• Osteonecrosis of the jaw (ONJ) presents complex clinical challenges, especially in resource-limited settings.
• Accurate diagnosis and timely treatment are crucial for effective ONJ management.

Purpose of the Study:

• To develop an integrated vision-language system (ONJ-TS) for supporting ONJ diagnosis and treatment.
• To enhance clinical decision-making and efficiency in managing ONJ patients.

Main Methods:

• Retrospective collection of panoramic radiography (PR) images and medical records from 366 ONJ patients.
• Fine-tuning the Qwen2-VL model for natural language question answering on ONJ treatment.
• Training SwinTransformer and U-Net models for ONJ case identification and region segmentation in PR images.

Main Results:

• The ONJ-TS system demonstrated high accuracy in diagnosis and treatment support.
• Human evaluation showed ONJ-TS accuracy exceeding junior oral and maxillofacial surgeons (OMFSs).
• Comparative analysis revealed ONJ-TS efficacy comparable to senior OMFSs in disease management.

Conclusions:

• The ONJ-TS system offers a promising digital solution for ONJ clinical decision support.
• The system has the potential to enhance clinical efficiency and treatment accuracy for ONJ.
• AI integration in medical imaging and language processing can significantly aid in managing complex conditions like ONJ.