Gait Parameters Classification Of Parkinson’s Disease Patients Based On Machine Learning

Parkinson’s disease（PD）has become the second most common neurodegenerative disease in the middle-aged and elderly people.With the aging of the population in my country,more and more people are affected by PD.PD patients mainly present with tremor,bradykinesia,gait balance disorders and other motor symptoms.The aggravation of the disease will further affect the quality of life of the patient,and even lead to the inability to take care of themselves.However,most PD patients are diagnosed by a unified Parkinson’s score scale,patient complaints,and physician clinical experience currently.Due to the different clinical experiences of each physician and the patient’s susceptibility to the environment,emotion,and other factors,this diagnosis method is highly subjective,and it is easy to miss diagnosis and misdiagnosis.Moreover,in recent years,there has been a high incidence and high growth trend of PD,and in this case,there is an urgent need for an effective means of monitoring and preventing the disease.Therefore,the purpose of this article is to explore an objective and reliable method for the clinically assisted diagnosis of PD.In recent years,with the rapid development of mechatronic systems and sensing technology,the usage of wearable sensors to collect the gait information of patients’ walking process for quantitative analysis has become a research focus in the field of PD.Therefore,based on the abnormal gait disorder exhibited by patients with PD,this thesis uses GYENNO MATRIX,a small-size,high-precision wearable motion evaluation device developed by GYENNO,to collect the gait information during walking.Finally,a classification model for identifying PD and normal gait is established by using machine learning methods based on gait characteristics.In this thesis,the design and implementation of the gait information acquisition system,the processing of gait features,the establishment of the gait feature classification model,and the evaluation of the classification model are studied,and the specific work content and results are as follows:（1）Constructs an i TUG gait acquisition system based on GYENNO MATRIX to collect the gait characteristics related to basic parameters,kinematic parameters and kinetic parameters of subjects during walking.（2）The collected gait features are preprocessed and feature selected,and then the classification model is established using traditional machine learning algorithms,including support vector machines（SVM）,decision trees（DT）,and logistic regression（LR）.（3）This thesis proposes a capsule network model to construct a classifier,and uses the traditional convolutional neural network for comparison,which reflects the superiority of the capsule network.In addition,compared with SVM,DT,and LR,the capsule network showed that it had better classification performance for PD patients and normal people.Experimental results: After the comparison of multiple classification models,the capsule network showed the best classification performance.Its accuracy,accuracy,recall rate and F1-score are 94.2%,95.05%,96.97% and 0.96,respectively.The results show that the classification model based on machine learning for the gait classification of PD patients and normal people has objectivity and reliability,and can effectively assist clinicians in the diagnosis of Parkinson’s disease,and it breaks the subjectivity brought about by traditional diagnostic methods.