Adaboost Based Gait Analysis And Its Application To Diagnosis Of Knee Osteo Arthritis

In China,the modern society,aged population has become a more and more serious problem.The morbidity of Knee Osteo Arthritis(KOA)has increased,so early diagnosis of potential KOA patients is of great significance for treatments.At present,KOA is usually diagnosed by imaging tests and Symptomatology,and relevant medical experts need to make judgments and decisions through their own experience and thoughts.The introduction of gait analysis and machine learning methods are not only enable a computer to update its own decision-making system by learning the prior knowledge of experts,but also assist relevant physicians to make diagnoses of KOA patients,improve the efficiency of related work as well as reduce the risks brought by patients missing the best treatment opportunity.Specific research contents are as follows.(1)Aiming at the problems of subjective factors and low efficiency in traditional manual methods commonly used in KOA,this research uses Gait Watch to conduct data acquisitions and gait analysis.By statistical analysis and correlation analysis of gait characteristics of 650 samples from Guangdong Second Provincial General Hospital and137 samples from intelligent digital laboratories,and gait analysis and evaluations between KOA and non-KOA patients,it is found that there is striking difference in gait characteristics between them,and the sample data set can be used for training and testing diagnostic classifiers.(2)Aiming at the inefficiency of gait analysis in KOA Diagnosis,k-nearest Neighbor,Logistic Regression,Naive Bayes and Support Vector Machine are applied to the training of diagnostic classifiers after the data set is preprocessed by the abnormal value processing and the normalization method.Each diagnosis classifier is evaluated and analyzed based on the classifier’s classification performance metrics,including Accuracy,Sensitivity,Specificity and AUC value.The results show that the diagnostic classifier of Radial Basis Function Support Vector Machine has the best classification performance after parameter optimization,but the overall classification performance stillhas room for improvement.(3)In order to solve the problem of insufficient classification performance of diagnostic classifiers,this research uses Ada Boost integrated learning methods to enhance the classification performance of diagnostic classifiers.By adopting Ada Boost integrated learning methods,the results showed that each diagnostic classifier had notable improvement in classification performance,among which the best was Radial Basis Function Support Vector Machine diagnostic classifier,obtaining accuracy of93.43 %,sensitivity of 0.97,specificity of 0.91 and AUC value of 0.9327.All the above results prove that Ada Boost based gait analysis and its application to diagnosis of K OA have great research value and significance.