Algorithm Research And System Implementation Of In Telligent Dental Lesion Detection

With the development of deep learning and the widespread use of medical imaging technology,the research of automatic detection of medical images through deep learning algorithms has gradually developed into a popular research direction.As an important branch of medical research,dentistry is mainly diagnosed based on dental X-rays.The traditional manual diagnosis method requires employees to have rich professional knowledge,and the diagnosis process is time-consuming and labor-consuming.In order to alleviate the diagnostic pressure of dentists and realize the intelligent diagnosis of dental lesions,this thesis studies the intelligent detection algorithm of dental lesions based on deep learning and its system implementation.The specific work includes three parts:1.Using convolution neural network to realize the intelligent diagnosis of dental lesions.Aiming at some common dental lesions such as periapical periodontitis,dental caries and apical cyst,a variety of convolution neural networks with different depths are used to train and test on self-made multi classification dataset,and the best model is selected through the comparison results;at the same time,in order to verify the accuracy of the extracted features,the visualization experiment is carried out;the results show that the DenseNet-121 network model used in this thesis has great advantages in recognition accuracy and speed,and can accurately identify the characteristics of each lesion type.2.A regional localization method of dental disease based on YOLOv5 is put forward for making the detection more accurate.Three algorithms,YOLOv5,SSD and Fast R-CNN,are used to train and test on the self-made caries data set,periapical lesion data set and combined lesion data set respectively.By comparing and analyzing test result,we get conclusion that the dental lesion region location method based on YOLOv5 proposed in this thesis has higher recognition accuracy,recall rate and precision rate compared with the SSD and Fast R-CNN algorithms used in the existing research.3.Based on the lesion classification algorithm and lesion area location algorithm,this thesis develops a set of dental intelligent diagnosis and treatment software which consists of three modules:user management module,intelligent diagnosis and treatment module and system home page module.In addition to completing the core lesion detection function,it also has certain data statistics function.