Research On Classification Method Of Rheumatoid Metacarpophalangeal Joint Ultrasound Image

With the explosive growth of medical imaging data,clinical diagnosis physicians are facing increasing work pressure,and a large number of ultrasound images of rheumatoid arthritis are needed to observe during the diagnosis process.In addition,the theoretical level of the clinician,clinical practice experience,and personal subjective judgment will all have an impact on the diagnosis.Therefore,the use of machine learning algorithms to establish a computer-aided diagnosis system for rheumatoid arthritis,assisting physicians in grading rheumatoid arthritis ultrasound images,can improve the efficiency of physicians’diagnosis and treatment,and shorten the period of patient visits,which has important research value and reality.significance.This paper studies the classification method of metacarpophalangeal joint ultrasound images of rheumatoid arthritis based on least squares support vector machine and random forest algorithm.The main works are as follows:1.A classification method of rheumatoid arthritis ultrasound images based on Least Square Support Vector Machine is presented.This method first preprocesses the ultrasound images of rheumatoid arthritis metacarpal arthritis,including ultrasound image denoising,image enhancement,and data augmentation.Then,according to the characteristics of the diseased area of the joint tissue,the Region of Interest(ROI)is selected,and the ROI image training data set and the test data set are constructed based on the statistical feature value of the gray histogram.Finally,use the Least Squares Support Vector Machine to train the classification model,and use the test data set to test the classification model.A large number of two-class and four-class experimental results further verify the classification method.2.A method for classification of ultrasound images of rheumatoid metacarpophalangeal joints based on Random Forest algorithm is studied.This method first extracts eigenvalues based on the gray-level co-occurrence matrix on the ultrasound image of the ROI area,and obtains a total of 88-dimensional eigenvalues in 4 angle directions.Then,based on the random forest algorithm,the classification of rheumatoid metacarpophalangeal joint ultrasound images was carried out based on the 88-dimensional feature vector,and the feature value combination with the lowest correlation was obtained through the mutual information feature selection method,so as to realize classification of ultrasound images.The outcomes indicate that in four-classification test,precision accuracy rate attains more than 90%when surpass 20 features are selected.3.Designed and implemented the software for the ultrasonic image-assisted diagnosis system of rheumatoid metacarpophalangeal joints,which can realize the functions of ultrasonic image preprocessing,ROI image feature extraction,ROI image display and storage,and diagnosis judgment result output.