~口腔疾病诸如牙结石、龋齿及牙龈炎等在我国患病率较高。开发口腔疾病智能诊断系统对于口腔疾病的提前发现、及时治疗具有重要作用，对于提高全民口腔健康水平具有重要意义。
首先构建口腔疾病数据集，使用Labelme标注软件对7220张口腔内图像进行标注，通过绘制多边形的方式标注图片并赋予不同的多边形不同的语义类别。之后对数据集进行预处理，将标注完成的原始图像和标注信息文件通过脚本批量处理为8位深度图，以供模型进行训练。使用9:1的训练策略，训练数据集占90%即6498张，测试数据集占10%即722张。
然后，口腔疾病的识别诊断需要定位病灶并对病灶进行分类，而将语义分割作为病灶的识别方式可以达到定位和分类的目的。研究了近年以来的各种语义分割模型的网络结构和基本原理，选取了三种基于编码器-解码器原理的语义分割模型作为理论指导。因此，基于U-Net、PSPNet与DeepLabV3+构建了三种的口腔疾病病灶识别模型。三种模型依次使用训练数据集进行训练，经过50个世代的训练，损失函数收敛。结果表明，基于U-Net的模型相较于其他两种模型识别效果更加精确，mIou和mPA评价指标更高。
其次，深入研究了注意力机制等卷积神经网络的改进办法，归纳总结了不同注意力机制的原理和作用并对模型进行重构。基于常见的通道和空间注意力机制，使用了SENet、CBAM、ECANet三种注意力模块。搭建了DR-UNet、CS-UNet和FD-UNet三种基于注意力机制的口腔疾病病灶分割模型。三种网络使用相同的训练数据集进行训练，经过80个epoch的迭代，损失函数收敛，改进后的网络模型损失函数有一定下降，并在mIou和mPA两项评价指标下都有一定的提升，其中CS-UNet效果最佳。
基于上述的研究，设计并实现了一套B/S架构口腔疾病诊断系统。该系统具有登陆和注销功能、基于图像识别的诊断功能和用户管理功能。管理员用户可对系统及普通用户进行管理，而普通用户可以使用平台的诊断功能。将深度学习模型集成入后台，执行图像的识别，后台数据库会记录诊断历史。

~Oral diseases such as dental calculus, dental caries and gingivitis have a high prevalence in China. The development of oral disease intelligent diagnosis system plays an important role in the early detection and timely treatment of oral diseases, and is of great significance to improve the oral health level of the whole nation.
Labelme labeling software was used to label 7220 oral images. The images were labeled by drawing polygons and given different semantic categories to different polygons. After that, the data set was preprocessed, and the original images and annotation information files were processed into 8-bit depth maps through scripts in batches for model training. Using the 9:1 training strategy, the training data set accounted for 90% (6498 pieces), and the test data set accounted for 10% (722 pieces).
Then, the identification and diagnosis of oral diseases need to locate and classify the lesions, and semantic segmentation can be used to identify the lesions. The network structure and basic principle of various semantic segmentation models in recent years are studied, and three semantic segmentation models based on encoder-decoder principle are selected as theoretical guidance. Therefore, based on U-NET, PSPNet and DeepLabV3+, three types of oral disease lesion recognition models were constructed. After 50 generations of training, the loss function converges. The results show that compared with the other two models, the recognition effect of U-NET model is more accurate, and mIou and mPA evaluation indexes are higher.
Secondly, the improvement methods of convolutional neural network such as attention mechanism are studied in depth. The principles and functions of different attention mechanisms are summarized and the model is reconstructed. Based on the common channel and spatial attention mechanism, three kinds of attention modules, SENet, CBAM and ECANet, are used. Three models of focus segmentation of oral diseases based on DR-UNet, CS-UNET and FD-UNET were established. The three networks used the same training data set for training. After 80 iterations of epoch, the loss function converges, and the loss function of the improved network model has decreased to some extent and improved to some extent under mIou and mPA, among which CS-UNET has the best effect.
Based on the above research, a B/S architecture oral disease diagnosis system was designed and implemented. The system has functions of login and logout, diagnosis and user management based on image recognition. Administrators can manage the system and common users, and common users can use the diagnostic function of the platform. The deep learning model is integrated into the background to perform image recognition, and the background database will record the diagnosis history.

[1] LeCun Y, Bengio Y, Hinton G. Deep learning[J]. nature, 2015, 521(7553): 436-444.

[2] Kim I, Baek W, Kim S. Spatially attentive output layer for image classification[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020: 9533-9542.

[3] Liu W, Anguelov D, Erhan D, et al. Ssd: Single shot multibox detector[C]//European conference on computer vision. Springer, Cham, 2016: 21-37.

[4] Redmon J, Divvala S, Girshick R, et al. You only look once: Unified, real-time object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 779-788.

[5] Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation[C]//International Conference on Medical image computing and computer-assisted intervention. Springer, Cham, 2015: 234-241.

[6] Zhao H, Shi J, Qi X, et al. Pyramid scene parsing network[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 2881-2890.

[7] Florian L C, Adam S H. Rethinking atrous convolution for semantic image segmentation[C]//Conference on Computer Vision and Pattern Recognition (CVPR). IEEE/CVF. 2017.

[8] Lee S, Kim J, Shin Yoon J, et al. Vpgnet: Vanishing point guided network for lane and road marking detection and recognition[C]//Proceedings of the IEEE international conference on computer vision. 2017: 1947-1955.

[9] He B, Li J, Zhao Y, et al. Part-regularized near-duplicate vehicle re-identification[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2019: 3997-4005.

[10] Huang L, An R, Zhao S, et al. A deep learning-based robust change detection approach for very high resolution remotely sensed images with multiple features[J]. Remote Sensing, 2020, 12(9): 1441.

[11] Rana A, Yauney G, Wong L C, et al. Automated segmentation of gingival diseases from oral images[C]//2017 IEEE Healthcare Innovations and Point of Care Technologies (HI-POCT). IEEE, 2017: 144-147.

[12] JD González, Rojas J Q . Use of Convolutional Neural Networks in Smartphones for the Identification of Oral Diseases Using a Small Dataset[J]. Revista Facultad de Ingeniería, 2021, 30(55):e11846.

[13] 郭飞飞,赖新鑫,汪璐璐.驻新疆某部官兵口腔健康及其认知情况调查[J].人民军医,2020,63(1):12-14,33.

[14] 李勇艳.1410例中老年人牙周疾病的患病情况及患病相关因素分析[J].当代医药论丛,2020,18(16):92-93.

[15] 荆进,李娜,刘鑫.牙周病患者口腔保健知识与个人口腔保健行为情况调研[J].国际护理学杂志,2020,39(15):2691-2694.

[16] 肖滢,刘怡然,沈红,仇颖莹,沈家平.江苏省12～15岁中学生牙周健康状况抽样调查[J].口腔医学,2020,40(03):249-254.

[17] 李素华,热依达·阿斯哈提,热西旦·扎克尔,褚雪倩,李冲,郭欣,帕里旦·吉拉,周小佳.新疆墨玉县农村中青年人群牙周现状及牙周炎危险因素分析[J].新疆医科大学学报,2020,43(05):647-652.

[18] 迟帅.沈阳地区老年人牙周病和龋病患病情况调查分析[J].中国卫生统计,2019,36(4):538-540.

[19] 刘洋主编. 口腔综合考研考点突破图解[M]. 2019

[20] Liu X, Faes L, Kale A U, et al. A comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging: a systematic review and meta-analysis[J]. The lancet digital health, 2019, 1(6): e271-e297.

[21] Lee J H, Kim D H, Jeong S N, et al. Detection and diagnosis of dental caries using a deep learning-based convolutional neural network algorithm[J]. Journal of dentistry, 2018, 77: 106-111.

[22] Prajapati S A, Nagaraj R, Mitra S. Classification of dental diseases using CNN and transfer learning[C]//2017 5th International Symposium on Computational and Business Intelligence (ISCBI). IEEE, 2017: 70-74.

[23] Casalegno F, Newton T, Daher R, et al. Caries detection with near-infrared transillumination using deep learning[J]. Journal of dental research, 2019, 98(11): 1227-1233.

[24] Chang H J, Lee S J, Yong T H, et al. Deep learning hybrid method to automatically diagnose periodontal bone loss and stage periodontitis[J]. Scientific reports, 2020, 10(1): 1-8.

[25] Pentapati K C, Mishra P, Damania M, et al. Reliability of intra-oral camera using teledentistry in screening of oral diseases–Pilot study[J]. The Saudi dental journal, 2017, 29(2): 74-77.

[26] Rana A, Yauney G, Wong L C, et al. Automated segmentation of gingival diseases from oral images[C]//2017 IEEE Healthcare Innovations and Point of Care Technologies (HI-POCT). IEEE, 2017: 144-147.

[27] Accuracy of remote diagnoses using intraoral scans captured in approximate true color: a pilot and validation study in teledentistry.[J]. BMC oral health, 2020, 20(1) : 266-266.

[28] Balaei Asghar Tabatabaei et al. Automatic detection of periodontitis using intra-oral images.[J]. Conference proceedings : ... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual Conference, 2017, 2017 : 3906-3909.

[29] 上海视可电子科技有限公司.一种牙病诊断方法、诊断设备及智能图像采集装置:CN201810403336.0[P].2018-08-09.

[30] 赵睿,张钊,刘立,甘中学,贾婷婷,冯杭,朱亮,乔波,邢乐君,张海钟,吴子忠.一种自助式口腔疾病早期诊断及口腔洁治机器人系统设计[J].海南医学,2018,29(23):3348-3350.

[31] 王飞,汪鹏,周琳,陈蜀虎,吴昊.基于深度学习的影像智能诊断平台建设与应用[J].中国数字医学,2020,15(01):11-13.

[32] 孟凡皓,田瑜,乔波, 等.基于大规模临床数据深度学习的口腔疾病人工智能预防与诊断平台的构建[J].精准医学杂志,2020,35(6):497-500.

[33] 范德增,单志鑫.我国口腔医疗行业市场现状与趋势分析[J].新材料产业,2019(08):3

[34] 王兴.我国口腔医疗行此的现状、机遇与挑战[J].对话资本,2018(5):5—8.

[35] 高学军. 牙体牙髓病学[M]. 北京大学医学出版社, 2013.

[36] 孟焕新. 牙周病学[M]. 人民卫生出版社, 2008.

[37] LeCun Y, Bengio Y, Hinton G. Deep learning[J]. nature, 2015, 521(7553): 436-444.

[38] 徐姗姗.卷积神经网络的研究与应用[D].南京：南京林业大学,2013.

[39] HUBEL D H, WIESEL T N. Receptive fields, binocular interaction and functional architecture in the cat's visual cortex[J]. The Journal of physiology, 1962, 160(1):106-154.

[40] HUBEL D H, WIESEL T N. Brain and visual perception: The story of a 25-year collaboration[M]. Oxford, England, United Kingdom:Oxford University Press, 2004.

[41] Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[J]. Advances in neural information processing systems, 2012, 25.

[42] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. arXiv preprint arXiv:1409.1556, 2014.

[43] Lin M, Chen Q, Yan S. Network in network[J]. arXiv preprint arXiv:1312.4400, 2013.

[44] He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 770-778.

[45] Szegedy C, Vanhoucke V, Ioffe S, et al. Rethinking the inception architecture for computer vision[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 2818-2826.

[46] Howard A G, Zhu M, Chen B, et al. Mobilenets: Efficient convolutional neural networks for mobile vision applications[J]. arXiv preprint arXiv:1704.04861, 2017.

[47] Zoph B, Vasudevan V, Shlens J, et al. Learning transferable architectures for scalable image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 8697-8710.

[48] Gastaldi X. Shake-shake regularization[J]. arXiv preprint arXiv:1705.07485, 2017.

[49] Zeng X, Ouyang W, Yan J, et al. Crafting gbd-net for object detection[J]. IEEE transactions on pattern analysis and machine intelligence, 2017, 40(9): 2109-2123.

[50] Google Brain. TensorFlow[EB/OL]. https://tensorflow.google.cn, 2019.

[51] Facebook.PyTorch[EB/OL].https://pytorch.org,2019

[52] Yangqing Jia. Caffe[EB/OL]. http://caffe.berkeleyvision.org, 2019.

[53] Long J, Shelhamer E, Darrell T. Fully convolutional networks for semantic segmentation[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2015: 3431-3440.

[54] Badrinarayanan V, Kendall A, Cipolla R. Segnet: A deep convolutional encoder-decoder architecture for image segmentation[J]. IEEE transactions on pattern analysis and machine intelligence, 2017, 39(12): 2481-2495.

[55] Chen L C, Papandreou G, Kokkinos I, et al. Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs[J]. IEEE transactions on pattern analysis and machine intelligence, 2017, 40(4): 834-848.

[56] Zhao H, Shi J, Qi X, et al. Pyramid scene parsing network[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 2881-2890.

[57] Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation[C]//International Conference on Medical image computing and computer-assisted intervention. Springer, Cham, 2015: 234-241.

[58] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. arXiv preprint arXiv:1409.1556, 2014.

[59] 谢飞,穆昱,管子玉,沈雪敏,许鹏飞,王和旭.基于具有空间注意力机制的Mask R-CNN的口腔白斑分割[J].西北大学学报(自然科学版),2020,50(01):9-15.

[60] Goyal P, Dollar P, Girshick R B, et al. Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour.[J]. arXiv: Computer Vision and Pattern Recognition, 2017.

[61] Smith S L, Le Q V. A bayesian perspective on generalization and stochastic gradient descent[J]. arXiv preprint arXiv:1710.06451, 2017.

[62] He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 770-778.

[63] Chen L C, Zhu Y, Papandreou G, et al. Encoder-decoder with atrous separable convolution for semantic image segmentation[C]//Proceedings of the European conference on computer vision (ECCV). 2018: 801-818.

[64] Chollet F. Xception: Deep learning with depthwise separable convolutions[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 1251-1258.

[65] Hu J, Shen L, Sun G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 7132-7141.

[66] Woo S, Park J, Lee J Y, et al. Cbam: Convolutional block attention module[C]//Proceedings of the European conference on computer vision (ECCV). 2018: 3-19.

[67] Wang Q , Wu B , Zhu P , et al. ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks[C]// 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2020.

[68] Lin G , Milan A , Shen C , et al. RefineNet: Multi-path Refinement Networks for High-Resolution Semantic Segmentation[J]. IEEE, 2017.

[69] Chen L C , Papandreou G , Kokkinos I , et al. Semantic Image Segmentation with Deep Convolutional Nets and Fully Connected CRFs[J]. Computer Science, 2014(4):357-361.

[70] Chen L C , Papandreou G , Kokkinos I , et al. DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40(4):834-848.