Research On The Diagnosis Of Parkinson’s Tremor Signal Based On Video

In medicine,some common diseases(such as Parkinson’s disease,stroke and epilepsy)could cause spontaneous tremors in patients,and doctors could make a preliminary diagnosis of the condition based on the tremor signals of different body parts of the patient.As one of the most common tremor diseases,Parkinson’s disease had an increasing incidence both at home and abroad in recent years.The extraction and analysis of Parkinson’s tremor signals had also become a hot spot in the research of Parkinson’s disease.Existing methods for extracting tremor signals generally required patients to wear related equipment.Although these devices could extract tremor signals more effectively,they were often expensive,complicated to wear,and not universal.In addition,clinical evaluation of Parkinson’s patients required doctors to score various parts of the patient according to UPDRS(Unified Parkinson’s Disease Rating Scale).This process required a lot of time and workload,which undoubtedly brings a lot of work to the medical unit,also it caused a huge pressure for medical treatment.This paper proposed a method of extracting Parkinson’s tremor signal from video and analyzing the tremor signal with the aid of a computer for the first time.The method first used Open Pose to obtain the position information of different body parts of the patient,and then took the position information as the center and took an appropriate range to intercept partial videos of different body parts.Since the original Open Pose network structure model was more complex and required a large amount of calculation,this paper proposed a network model with less calculation amount.After obtaining local videos of different body parts of patients with tremor disease,the Euler video amplification algorithm was used to extract the initial tremor signal from the local video.In order to get better results,the Butterworth band-pass filter was designed to filter the tremor signal according to the frequency band range of the Parkinson’s tremor signal.The experimental in this paper took videos of 75 Parkinson’s disease patients as experimental subjects.Through the method proposed in this paper,partial videos of different body parts of patients were extracted and tremor signals were extracted from the videos.Then we used the designed Butterworth band-pass filter to filter and then use BP neural network and one-dimensional convolutional neural network to predict the score of UPDRS for the obtained signals.The final experimental results showed that the method proposed in this paper could effectively extract the Parkinson’s tremor signal and predicted the UPDRS score.In the future,this method could be considered to assist doctors in the preliminary assessment of Parkinson’s disease patients,which would greatly reduce the outpatient pressure of the medical department.In addition,the proposed method for analyzing the characteristics of Parkinson’s tremor signals could effectively extract the important features of Parkinson’s tremor signals,and the experimental results showed that these features had a great correlation with the condition of Parkinson’s disease patients,which had a good research value for the diagnosis of Parkinson’s disease patients.