Tooth Segmentation Using SAM2 Model in Dental X-Ray Images
Author Information
Author(s): Li Zifeng, Tang Wenzhong, Gao Shijun, Wang Yanyang, Wang Shuai
Primary Institution: Beihang University
Hypothesis
Can the SAM2 model be adapted for effective tooth segmentation in dental panoramic X-ray images?
Conclusion
The S2AgScUNet model significantly improves tooth segmentation accuracy and efficiency compared to traditional models.
Supporting Evidence
- The S2AgScUNet model achieved an IoU score of 0.8612, surpassing the traditional UNet model.
- LightUNet, a lightweight model, maintained performance comparable to UNet while significantly reducing parameters.
- Knowledge distillation was effectively used to create a model suitable for edge devices.
Takeaway
This study shows a new way to help dentists quickly and accurately see teeth in X-ray pictures, making it easier to find problems.
Methodology
The study used a fine-tuning approach with the pre-trained SAM2 model, incorporating adapter modules and knowledge distillation to create a lightweight model for deployment.
Limitations
The study is limited by the small size of the dental X-ray dataset, which may affect the generalizability of the results.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website