Scholarlink

<p>Automatic and early recognition of dental caries using machine vision is essential due to the increasing prevalence of tooth decay. Early caries detection can lower treatment costs, preserve tooth structure, and avoid related health problems. Even though artificial intelligence has come a long way recently, existing dental datasets often have limited diversity, quantity, and annotation quality. With this aim, this paper introduces a comprehensive self-built dental caries dataset, providing a solid basis for training caries recognition models. Our caries dataset contains about 4,700 images collected from publicly available sources before being carefully labeled and validated by dental experts. To establish robust benchmarks, we also conduct extensive experiments using state-of-the-art machine vision algorithms, including YOLO-based models, Faster R-CNN, and Mask R-CNN. The experiments show that the self-built dataset works well; the mean average precision (mAP50:95) scores were over 53%, and the real-time processing speeds of less than 30&#xa0;ms per image using YOLO-based models. CariXray is proposed to help with advanced research and real-world use of computer-aided dental diagnostics.</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

Carixray: a periapical X-ray dataset for machine vision-based dental caries recognition

  • 期刊论文

  • 发表日期：

    2025年12月29日

查看全文
  • Tuan Linh Dang,
  • Trong Nghia Nguyen,
  • Tuan Minh Vu

摘要

Automatic and early recognition of dental caries using machine vision is essential due to the increasing prevalence of tooth decay. Early caries detection can lower treatment costs, preserve tooth structure, and avoid related health problems. Even though artificial intelligence has come a long way recently, existing dental datasets often have limited diversity, quantity, and annotation quality. With this aim, this paper introduces a comprehensive self-built dental caries dataset, providing a solid basis for training caries recognition models. Our caries dataset contains about 4,700 images collected from publicly available sources before being carefully labeled and validated by dental experts. To establish robust benchmarks, we also conduct extensive experiments using state-of-the-art machine vision algorithms, including YOLO-based models, Faster R-CNN, and Mask R-CNN. The experiments show that the self-built dataset works well; the mean average precision (mAP50:95) scores were over 53%, and the real-time processing speeds of less than 30 ms per image using YOLO-based models. CariXray is proposed to help with advanced research and real-world use of computer-aided dental diagnostics.