Research On Intelligent Diagnosis Algorithm Of Dental Diseases Based On X-ray Film

The rapid development of artificial intelligence is driving medical reform,and deep learning has shown important practical application value in the field of medical image processing.The people’s emphasis on oral health has promoted the digitalization of dental care.However,in recent years,affected by the pandemic,patients with oral diseases still have a fear of visiting hospitals in the field for medical treatment,which delays the best time to treat dental diseases.In order to facilitate the diagnosis of dental disease patients in nearby clinics,avoid the intensive gathering of patients and relieve the pressure of dentists’ diagnosis,this paper carries out the research of intelligent diagnosis algorithm of dental disease based on X-ray film,and the specific work is divided into the following three parts.1.For a variety of common dental diseases such as dental caries and root disease,this paper proposes to use convolutional neural network to achieve intelligent diagnosis of dental diseases.The data set is established in cooperation with professional physicians.different convolutional neural networks are used for training and testing,and the best network model is selected for preliminary diagnosis through comparison experiments.The experimental results show that the MobileNetV3 network has high performance and can identify various lesion images more accurately.2.To address the problem of similar periodontal disease performance,this paper proposes a Mask R-CNN network-based lesion region segmentation algorithm.Firstly,we use multiple network models to design comparison experiments,and select the better performing Mask R-CNN as the base network model for improvement.Referring to the characteristics of YOLOv5 model,the category target box prediction branch of Head module in the network model was replaced,and faster computing speed,higher segmentation accuracy and smaller model weights were obtained.The segmentation results were further fine-tuned by using the Marked Watershed algorithm to better fit the physician labeled images and obtain more accurate periodontal tissue lesion areas.3.To address the problem of difficult diagnosis through lesion images,a neural network diagnosis model based on texture features is designed in this paper.Firstly,the texture features of the lesion area were extracted by using gray scale co-generation matrix and Tamura texture,and the distribution of the feature statistics among different periodontal diseases was analyzed by combining with the radiographic performance.After that,multiple diagnostic models based on texture features were established to compare the diagnostic results.The experiments showed that the neural network model based on texture features could diagnose periodontal diseases more accurately.