Scholarlink

<p>This study introduces a robust framework for automated tooth detection and segmentation in panoramic dental radiographs (OPGs) to overcome challenges like limited data availability and anatomical variations. We curated and publicly released an 800- image, annotated OPG dataset that was augmented fourfold to 3200 images for enhanced model general- ization. The core of our methodology involves two novel two-stage hybrid pipelines: (1) vision transformer for region proposal with U-Net refinement, and (2) YOLOv8 detection combined with Double U-Net segmentation. YOLOv8 + Double U-Net pipeline achieved state-of-the- art performance with an intersection over union (IoU) of 0.91, significantly surpassing previous methods. This framework offers a clinically relevant solution that inte- grates efficient object detection with precise pixel-level segmentation, establishing a valuable decision-support tool for dental practitioners.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Teeth in focus: a deep dive into convolutional neural networks for dental image analysis

  • 期刊论文

  • 发表日期：

    2025年11月17日

查看全文
  • Vipul Arya,
  • Sathya Duraiswamy,
  • D. Jagadeesan

摘要

This study introduces a robust framework for automated tooth detection and segmentation in panoramic dental radiographs (OPGs) to overcome challenges like limited data availability and anatomical variations. We curated and publicly released an 800- image, annotated OPG dataset that was augmented fourfold to 3200 images for enhanced model general- ization. The core of our methodology involves two novel two-stage hybrid pipelines: (1) vision transformer for region proposal with U-Net refinement, and (2) YOLOv8 detection combined with Double U-Net segmentation. YOLOv8 + Double U-Net pipeline achieved state-of-the- art performance with an intersection over union (IoU) of 0.91, significantly surpassing previous methods. This framework offers a clinically relevant solution that inte- grates efficient object detection with precise pixel-level segmentation, establishing a valuable decision-support tool for dental practitioners.