Research On Computer Aided Diagnosis Of Fracture Based On X-ray Image

Purpose:Fracture is one of the most common disease types in human life,and it will cause serious harm to human health.If it is not possible to accurately diagnose the specific type of fracture,it will not only affect the subsequent treatment,but also delay fracture.Best treatment time,causing more serious consequences.Currently,one of the common fractures in clinical fractures is the distal radius fracture,and its common fracture type is checked as X-ray imaging.However,due to the unevenness of the fracture X-ray image itself,there is a characteristic of partial noise and artifacts,causing the doctor to easily produce deviation during the naked eye recognition of the fracture,which in turn affects its diagnostic accuracy.Therefore,the accurate classification task of the specific type of the distal radius fracture is used to achieve the specific type of the distal radius fracture.Methods:This experiment focuses on the detection and segmentation of the distal radius fracture lesions and the classification of the specific type of the distal radius fracture.The experiment first selects the contrast contrast adaptive histogram equalization algorithm（CLAHE）to improve the contrast of the fracture image main body portion and the background,to protrude the internal details of the fracture image so that the lesion information inside the fracture image is better.Subsequently,the Faster R-CNN target detection algorithm was selected to realize the detection task of the distal radius fracture lesion,and the feature extraction step in the Faster R-CNN original algorithm is improved,and the original feature extraction layer VGG-16 is replaced with a good effect.101,thereby further improving the detection performance of the Faster R-CNN algorithm.In terms of the classification task of the specific type of X-ray image of the humerus,this paper performs related experimental research from the characteristic extraction and classification of the image.The feature extraction portion mainly utilizes the traditional and depth features,and a new feature fusion method is proposed to better represent fracture images to improve the accuracy of the diagnosis of subsequent fractures.In the selection of the classifier,this paper selects the current classic machine learning classifier and depth learning classification network to achieve classification tasks of the distal radius fracture image.Results:The distal radius fracture diagnosis model constructed in this paper has achieved better classification effect,and the accuracy of Faster R-CNN detection is88.07%.Finally,the accuracy and F₁values of the distal radius fracture classification were 84.12%and 0.807.Respectively,and the method was proved by comparative experiments than the traditional fracture diagnosis has a higher classification accuracy.Conclusion:In this study,the target detection algorithm was used to locate the fracture site and the classification has achieved better classification results,which has a certain meaning of the distal radius fracture diagnosis.In the future,research also needs to improve the detection rate of fracture fine lesions,and further optimize the experimental design.