Research On Diagnosis Method Of Bone X-ray Image Based On Traditional Machine Vision Method

Many people in the world suffer from musculoskeletal diseases every year such as fractures and bone degeneration.Professional doctors need to diagnose the X-rays of the affected parts based on experience accumulated in lots of years.With the huge work pressure,low efficiency and less experience,The doctor may have a misdiagnosis.With the continuous development of technology,artificial intelligence technology has been applied to many fields of auxiliary medicine.In order to promote the development of automatic X-ray image diagnostic system,Stanford University released the world's largest bone X-ray data set in 2018(MURA data set).At present,the results published on this dataset are basically based on deep neural network-based classification methods.Based on this dataset,this paper attempts to use computer vision and classification methods and gives reference results.In addition,the original data only gives a skeletal abnormality in the X-ray film,but does not give accurate information such as the specific location and severity of the abnormality.Through the manual calibration work for a long time,this paper set a data set to detect abnormal bone position.The data set lays the foundation for further research work.The main work of this paper includes:(1)X-ray film classification method based on traditional computer visionFirstly,the original image is enhanced by image processing method,then the visual features of image such as LBP and HOG are extracted,and then the feature dimension reduction methods such as PCA and LDA are used to reduce dimension of image feature.Finally,the paper gives the classification results of RF,XGBoost and other classification methods by using different parameters.(2)Calibration of abnormal partsThe original MURA data set was calibrated by ways of my own detection and seeking expert help detection,and the abnormal position of the bone is accurately calibrated to form a new data set.The accurately calibrated data set lays a foundation for accurate diagnosis of skeletalabnormalities.(3)Method for detecting abnormal parts based on sliding windowBased on the accurate calibration of the data set in the abnormal part,this paper gives the detection results of the abnormal part of the skeleton based on the sliding window method,which has certain reference significance for the follow-up work.