Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

OralBBNet: Spatially Guided Dental Segmentation of Panoramic X-Rays with Bounding Box Priors

Created by
  • Haebom

Author

Devichand Budagam, Azamat Zhanatuly Imanbayev, Iskander Rafailovich Akhmetov, Aleksandr Sinitca, Sergey Antonov, Dmitrii Kaplun

Outline

This paper focuses on tooth segmentation and recognition, which play an important role in dental applications and diagnostic procedures. Although previous studies have addressed tooth segmentation, there are not many methods that successfully perform tooth segmentation and detection simultaneously. This study presents a new dental dataset called UFBA-425, which contains 425 panoramic dental X line images with bounding boxes and polygon annotations. In addition, we present OralBBNet architecture, which is designed to improve the accuracy and robustness of tooth classification and segmentation from panoramic X line images by combining the strengths of U-Net and YOLOv8. OralBBNet improves the mean accuracy (mAP) of tooth detection by 1-3% compared to existing techniques, improves the Dice score of tooth segmentation by 15-20% compared to state-of-the-art (SOTA) solutions, and improves the Dice score by 2-4% compared to other SOTA segmentation architectures. These results lay the foundation for the widespread implementation of object detection models in dental diagnosis.

Takeaways, Limitations

•
Takeaways:
◦
New dental dataset UFBA-425 available.
◦
We present OralBBNet architecture, which combines the strengths of U-Net and YOLOv8.
◦
Improved tooth detection and segmentation performance compared to existing technologies (improved mAP and Dice scores).
◦
Expanding the potential of object detection models in dental diagnosis.
•
Limitations:
◦
Lack of specific mention of the size and diversity of the UFBA-425 dataset.
◦
Further validation of the generalization performance of the OralBBNet architecture is needed.
◦
Lack of performance evaluation for various dental X line photograph types.
◦
A more comprehensive comparative analysis with other object detection and segmentation models is needed.
👍
Made with Slashpage