Scholarlink

Objective <p>To develop an artificial intelligence-based deep learning model for early screening and detection of periodontal diseases using periodontal images captured by a portable oral endoscope.</p> Methods <p>A dataset consisting of 811 clinical oral periodontal images collected via endoscopy was constructed. After data cleaning and high-quality annotation, it was divided into 271 healthy and 540 unhealthy periodontal images, with an 8:2 split for training and validation. The YOLOv11 deep learning model was used for analysis and early detection of periodontal diseases. Model performance was evaluated using precision, recall, and mean average precision.</p> Results <p>In the validation set, for binary classification the model achieved a precision of 0.727, a recall rate of 0.583, and a mean average precision of 0.632 at an IOU (Intersection over Union) threshold of 0.5. For multiclass detection, the model’s precision for identifying gingival recession was 0.671, for gingival swollen was 0.609, and for detecting dental calculus was 0.573.</p> Conclusion <p>The developed deep learning model enables low-cost early screening and detection of periodontal diseases, providing an important direction for the application of artificial intelligence in early periodontal disease screening.</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

AI-based early screening of periodontal diseases using endoscopic imaging

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月4日

查看全文
  • Ting Zhang,
  • Jinlei Yin,
  • Yufeng Wang,
  • Liangbo Li,
  • Nenghao Jin,
  • Liang Zhu,
  • Jian Wang,
  • Suixin Hu,
  • Lejun Xing,
  • Haizhong Zhang

摘要

Objective

To develop an artificial intelligence-based deep learning model for early screening and detection of periodontal diseases using periodontal images captured by a portable oral endoscope.

Methods

A dataset consisting of 811 clinical oral periodontal images collected via endoscopy was constructed. After data cleaning and high-quality annotation, it was divided into 271 healthy and 540 unhealthy periodontal images, with an 8:2 split for training and validation. The YOLOv11 deep learning model was used for analysis and early detection of periodontal diseases. Model performance was evaluated using precision, recall, and mean average precision.

Results

In the validation set, for binary classification the model achieved a precision of 0.727, a recall rate of 0.583, and a mean average precision of 0.632 at an IOU (Intersection over Union) threshold of 0.5. For multiclass detection, the model’s precision for identifying gingival recession was 0.671, for gingival swollen was 0.609, and for detecting dental calculus was 0.573.

Conclusion

The developed deep learning model enables low-cost early screening and detection of periodontal diseases, providing an important direction for the application of artificial intelligence in early periodontal disease screening.