Implementation Of Dental Caries Prevention Algorithms In Children

Dental caries is one of the three major diseases to be prevented and treated in the world,and childhood caries has attracted more attention in all age groups.Most of the children who go to the hospital are accompanied by their parents because of intolerable dental caries,and eventually have to bear the consequences of tooth extraction and filling,and face high treatment costs.At present,the available data sets and reference methods of children’s dental caries are relatively scarce,and conventional dental caries detection methods are difficult to achieve the early diagnosis of children’s dental caries.With the continuous maturity and improvement of deep learning technology and algorithm,it provides a new way and method for the early diagnosis of children’s dental caries.The aim of this study is to design a set of algorithms to automatically track and identify the caries of teeth,and give a solution to automatically track the changes of caries lesions in the high incidence of dental caries.The programme consists of three parts:The first part is to establish the maxillofacial morphological standards.According to the principles of medicine and tooth morphology,the maxillofacial feature standards of molars are constructed,including maxillofacial contour and fossa and groove morphology.There are 13 types of maxillofacial morphological standards of molars.The second part is to use U-net network and attention mechanism to realize dental and maxillofacial recognition and classification.U-net is used as the main network to classify and train the standard of maxillofacial contour and fossa and sulcus shape and compare the effects.U-net network with attention mechanism is introduced to compare and analyze,and then try to fuse the two morphological features.The effect of accurate classification of teeth was achieved,and the video stream detection was realized.The third part is to realize the early diagnosis of caries lesions by chalky algorithm detection.The image segmented by U-net network is input into the caries detection algorithm.The HSV range of chalky color obtained by the experiment is used for image processing,and the ratio of chalky pixels to maxillofacial pixels is taken as the judgment basis.The experimental results show that the accuracy of recognition after multi-model fusion is 0.7792,F1-score is 0.9061,and AUC is 0.9026.Compared with the original U-net network and based on the fossa and sulcus shape standard,the accuracy is increased by 15.78%,AUC is increased by 10.57%,and F1 score is increased by7.01%.Compared with the recognition based on maxillofacial contour morphological standard,the accuracy is increased by 3.8%,AUC is increased by 9.07%,and F1 score is increased by 10.15%.The completion of this study simplifies the complicated steps of traditional dental caries examination,realizes the flexible examination and accurate prevention of children’s dental caries in the field of home intelligent medical treatment,and reduces the rate of children’s caries.The experiments verify that the fusion of the two features has better recognition effect and the effectiveness of the attention mechanism.Finally,the dental and maxillofacial recognition model and caries lesion detection model were spliced by multi-model fusion technology,and the caries status of each molar was automatically tracked and diagnosed under the video stream.