南方医科大学学报

南方医科大学学报 ›› 2025, Vol. 45 ›› Issue (8): 1791-1799.doi: 10.12122/j.issn.1673-4254.2025.08.23

• • 上一篇    

YOLOv11-TDSP：口腔全景片的轻量化高精度异常牙检测模型

赵涛涛(), 倪铭(), 夏顺兴, 焦粤豪, 何亚婷   

  1. 四川农业大学信息工程学院，四川 雅安 625014
  • 收稿日期:2025-03-03 出版日期:2025-08-20 发布日期:2025-09-05
  • 通讯作者: 倪铭 E-mail:1425067966@qq.com;nm@sicau.edu.cn
  • 作者简介:赵涛涛，在读硕士研究生，E-mail: 1425067966@qq.com
  • 基金资助:
    四川省自然科学基金(2022NSFSC0172)

An lightweight algorithm for multi-dimensional optimization of intelligent detection of dental abnormalities on panoramic oral X-ray images

Taotao ZHAO(), Ming NI(), Shunxing XIA, Yuehao JIAO, Yating HE   

  1. College of Information Engineering, Sichuan Agricultural University, Ya'an 625014, China
  • Received:2025-03-03 Online:2025-08-20 Published:2025-09-05
  • Contact: Ming NI E-mail:1425067966@qq.com;nm@sicau.edu.cn

RichHTML

1

PDF

34

摘要：

目的 为解决口腔全景片中异常牙识别在目标检测方面的不足，通过对基准模型YOLOv11n进行优化，提出YOLOv11-TDSP模型。 方法 融合SHSA单头注意力机制与backbone层的C2PSA，形成新的C2PSA_SHSA注意力机制，通过部分输入通道运用单头注意力，削减计算冗余，提升模型效率与检测准确性。在head层增设小目标检测层，解决小目标易漏检、误检的难题。实施两次结构化剪枝，降低模型参数量，避免过拟合，提升平均精度。训练前期，采用亮度增强、伽马对比度调整等数据增强手段，增强模型泛化能力。 结果 优化后的YOLOv11-TDSP模型精度达94.5%、召回率92.3%、平均精度95.8%，相比基准模型YOLOv11n，分别提升6.9%、7.4%、5.6%。与高精度的YOLOv11x相比，参数量和计算量仅为其12%、13%。 结论 成功实现多种牙齿病症的轻量化、高精准识别。

关键词: 口腔全景片, 异常牙识别, 目标检测, YOLOv11n

Abstract:

Objective We propose a YOLOv11-TDSP model for improving the accuracy of dental abnormality detection on panoramic oral X-ray images. Methods The SHSA single-head attention mechanism was integrated with C2PSA in the backbone layer to construct a new C2PSA_SHSA attention mechanism. The computational redundancy was reduced by applying single-head attention to some input channels to enhance the efficiency and detection accuracy of the model. A small object detection layer was then introduced into the head layer to correct the easily missed and false detections of small objects. Two rounds of structured pruning were implemented to reduce the number of model parameters, avoid overfitting, and improve the average precision. Before training, data augmentation techniques such as brightness enhancement and gamma contrast adjustment were employed to enhance the generalization ability of the model. Results The experiment results showed that the optimized YOLOv11-TDSP model achieved an accuracy of 94.5%, a recall rate of 92.3%, and an average precision of 95.8% for detecting dental abnormalities. Compared with the baseline model YOLOv11n, these metrics were improved by 6.9%, 7.4%, and 5.6%, respectively. The number of parameters and computational cost of the YOLOv11-TDSP model were only 12% and 13% of those of the high-precision YOLOv11x model, respectively. Conclusion The lightweight YOLOv11-TDSP model is capable of highly accurate identification of various dental diseases on panoramic oral X-ray images.

Key words: panoramic oral X-ray images, abnormal teeth recognition, target detection, YOLOv11n