图像理解驱动的基牙选择多要素分析研究

Image Understanding-driven Multifactor Analysis for Abutment Tooth Selection

摘要

在可摘局部义齿(RPD)设计中,基牙的选择是关键,精准的基牙选择方法能为牙科医生提供重要的决策支持.然而,现有研究方法存在牙体微小病变识别灵敏度低、缺乏多维度综合分析等问题.为此,本文提出了一种面向基牙选择的多要素牙齿病症识别集成方法,设计了两种基于YOLOv11框架的改进模型:DSF-YOLOv11s-detect和HMS-YOLOv11s-pose.其中,DSF-YOLOv11s-detect模型聚焦于识别不适宜作为基牙的牙体状态,通过动态混合卷积网络(Dynamic Inception Mixer)和小物体增强金字塔(SOEP)提升牙齿特征提取能力,结合频域-空间注意力机制(FSA)和ShapeIoU损失函数,显著提高了口腔影像中小目标的识别精度.实验显示,该模型在牙位、种植牙、根管治疗未修复冠、桥体等检测任务中达到94.4%的mAP,帧率达155 frame/s.HMS-YOLOv11s-pose模型则通过关键点定位评估牙周炎分期,采用高频增强残差块(HFERB)和多尺度注意力机制(MSGA)优化关键点检测精度,结合SIoU损失函数增强方向敏感性.该模型在关键点检测中取得97.3%的mAP和93.6%的Macro-F1,帧率达185frame/s.两模型在对比实验中均优于主流方法,并在公开数据集DENTEX上实现精准可视化预测,进一步验证该多要素分析方法在辅助牙科医生进行RPD基牙选择智能诊断方面具有显著应用潜力和优势.

Abstract

In removable partial denture(RPD) design,the selection of abutment teeth is critical,and accurate abutment selection methods can provide important decision support for dentists.However,existing research methods have problems such as low sensitivity in identifying small lesions in the dentition and lack of comprehensive multi-dimensional analysis.To this end,this paper proposes an integrated method for multifactor dental lesion identification for abutment tooth selection,and designs two improved models based on the YOLOv11 framework:the DSF-YOLOv11s-detect and the HMS-YOLOv11s-pose.The DSF-YOLOv11s-detect model focuses on identifying dental states that are unsuitable for abutment teeth.The tooth feature extraction capability is enhanced by Dynamic Inception Mixer(DIM) and Small Object Enhancement Pyramid(SOEP),which,combined with Frequency Domain-Spatial Attention(FSA) mechanism and ShapeIoU loss function,significantly improves the accuracy of recognizing small targets in oral images.Experiments show that the model achieves 94.4% mAP with a frame rate of 155 frames/s in the detection tasks of tooth position,dental implant,root canal treatment of unprosthetic crowns,and bridges.The HMS-YOLOv11s-pose model,on the other hand,evaluates the periodontitis staging by keypoint localization,optimizes the keypoints by using the high-frequency-enhanced residual block(HFERB) and the multi-scalar attentional mechanism(MSGA) detection accuracy,combined with SIoU loss function to enhance orientation sensitivity.The model achieves 97.3% mAP and 93.6% Macro-F1 in keypoint detection with a frame rate of 185 frame/s.Both models outperform mainstream methods in comparative experiments and achieve accurate visualization prediction on the publicly available dataset DENTEX,which further validates that this multifactorial analysis method has significant application potential and advantages in assisting dentists to perform intelligent diagnosis of RPD abutment selection.significant application potential and advantages.

关键词

图像理解 / 可摘局部义齿基牙 / 智能诊断 / 口腔医学影像

Key words

image understanding / removable partial denture abutments / intelligent diagnosis / dental medical images

[1] Lee J,Kim Y,Park S,et al.Application of artificial intelligence in dentistry:current status and future perspectives[J].Journal of Artificial Intelligence in Dentistry,2025,1(2):30-39.
[2] Surdu A,Budala D G,Luchian I,et al.Using AI in optimizing oral and dental diagnoses—anarrative review[J].Diagnostics,2024,14(24):2804-2818.
[3] GUO M,ZHANG X H,LI Y.Integrated attentional teacher mutual consistency semi-supervised medical image segmentation[J].Computer Engineering,2024,50(9):313-323.
[4] FENG X P.Oral health status of Chinese residents—the fourth national oral health epidemiological survey report[C]//Proceedings of the 18th Annual Conference on Oral Preventive Medicine of Chinese Stomatology Society,2018:2-11.
[5] CHENG N,SHAN W X,WEN T,et al.Computer-aided design of implant-supported localized prosthetic for immediate repair of the mandibular defects type II[J].Chinese Journal of Stomatology,2024,59(11):1140-1144.
[6] National Health Commission.Guidelines for the diagnosis and treatment of dental caries,dental arch defects a nd missing teeth(2022 Edition)[Z].National Health Commission Office Medical Letter,[2022] ,339,2022.
[7] Park J,Lee J,Moon S,Lee K.Deep learning based detection of missing tooth regions for dental implant planning in panoramic radiographic images[J].Applied Sciences,2022,12(3):1595-1607.
[8] Xue T,Chen L,Sun Q.Deep learning method to automatically diagnose periodontal bone loss and periodontitis stage in dental panoramic radiograph[J].Journal of Dentistry,2024,150(6):105373-105386.
[9] ZENG X Q,XIA B,CAO Z Q,et al.Development of a recognition model for dental numerical abnormalities in children′s panoramic X-ray films based on deep learning[J].Chinese Journal of Stomatology,2023,58(11):1138-1144.
[10] LIANG H,QIU D Q,DING S Y,et al.Improved YOLOv8 for abnormal teeth and prosthesis detection in dental X-Ray imaging[J].Chinese Journal of Lasers,2025,52(3):94-104.
[11] Gorrell G,Alaa Daoud.Clinical guide to practical dental prosthodontics[M].Shenyang:Liaoning Science and Technology Publishing House,2023.
[12] Zhou Z,He A,Wu Y,et al.Spatial-frequency dual domain attention network for medical image segmentation[C]//IEEE International Conference on Bioinformatics and Biomedicine(BIBM),2024:4076-4081.
[13] Zhang H,Zhang S.Shape-iou:more accurate metric considering bounding box shape and scale[J].arXiv preprint arXiv:2312.17663,2023.
[14] Cui Y,Ren W,Knoll A,et al.Omni-kernel network for image restoration[C]//Proceedings of the AAAI Conference on Artificial Intelligence,2024:1426-1434.
[15] Li A,Zhang L,Liu Y,et al.Feature modulation transformer:cross-refinement of global representation via high-frequency prior for image super-resolution[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision,2023:12514-12524.
[16] Kolahi S G,Chaharsooghi S K,Khatibi T,et al.MSA MYM^2MYM net:multi-scale adaptive attention-guided network for medical image segmentation[J].arXiv preprint arXiv:2407.21640,2024.
[17] Gevorgyan Z.SIoU loss:more powerful learning for bounding box regression[J].arXiv preprint arXiv:2205.12740,2022.
[18] Yun S,Ro Y.Shvit:single-head vision transformer with memory efficient macro design[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,2024:5756-5767.
[19] Dai T,Wang J,Guo H,et al.FreqFormer:frequency-aware transformer for lightweight image super-resolution[C]//Proceedings of the International Joint Conference on Artificial Intelligence,2024:731-739.
[20] Shi D.Transnext:robust foveal visual perception for vision transformers[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,2024:17773-17783.
[21] Ma X,Dai X,Bai Y,et al.Rewrite the stars[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,2024:5694-5703.
附中文参考文献:
[3] 郭 敏,张熙涵,李 阳.融合注意力的教师互一致性半监督医学图像分割[J].计算机工程 ,2024,50(9):313-323.
[4] 冯希平.中国居民口腔健康状况—第4次中国口腔健康流行病学调查报告[C]//中华口腔医学会第18次口腔预防医学学术年会论文汇编,2018:2-11.
[5] 成 妮,单文汐,温 涛,等.计算机辅助设计种植体支持式可摘局部义齿即刻修复下颌Kennedy第2类牙列缺损1例[J].中华口腔医学杂志,2024,59(11):1140-1144.
[6] 国家卫生健康委员会.牙体缺损、牙列缺损与缺失修复诊疗指南(2022年版)[Z].国卫办医函〔2022〕339号,2022.
[9] 曾雪晴,夏 斌,曹战强,等.基于深度学习的儿童曲面体层X线片牙齿数目异常识别模型的研发[J].中华口腔医学杂志,2023,58(11):1138-1144.
[10] 梁 洪,邱定乾,丁世宇,等.基于改进YOLOv8的异常牙齿和修复体X射线影像检测[J].中国激光,2025,52(3):94-104.
[11] 戈登·格雷,阿拉·达乌德.实用牙体修复临床操作指南[M].沈阳:辽宁科学技术出版社,2023.

基金资助

辽宁省应用基础研究计划项目(2023JH2/101300016)资助;中国医科大学虚拟仿真建设项目(YDXF2024041)资助.