Research On The Algorithm Of Critical Event Detection In Parkinson's Disease

Parkinson's disease(PD)is a global,second-largest neurodegenerative brain disease,which seriously affects the health and life quality of patients.According to statistics,there are more than 10 million patients suffering from PD worldwide,with nearly 3 million in China.PD is expected to be found,diagnosed and treated at the early stage.Motor symptoms are important components in PD diagnosis.Clinically,tremor,bradykinesia,stiffness,hypokinesia,dyskinesias and gait impairments are the main motor symptoms of PD.At present,rating scale is mainly utilized to assess the motor symptoms.However,this method has a subjective limitation from the patient's complaint and the doctor's judgment.Tremor and Freezing of Gait(FOG)are PD typical motor symptoms.Automatic identification of tremor and FOG shows significant importance in clinical trials.This paper,which is based on wearable inertial sensor equipment,studied the identification of critical events of tremor and FOG,then explored the method of PD early detection based on the critical events identification.Based on the clinical scales and experimental thought,this study designed experiments for measuring tremor and FOG,and verified the results with database collected from a set of five-node 6-axis inertial sensors worn on wrist,shank and waist.The database included motor data of 110 experiments from 90 patients and 17 experiments from 17 healthy elders as control group.In the tremor recognition study,the experiments are designed in standing position and sitting position.According to the conventional research,PD characteristic tremor is described as an approximate sinusoidal motion of 4~7hz.In this paper,the Brug's AR method power spectral density(PSD)is selected to calculate the frequence,and a root mean square(RMS)criterion was used to ensure the frequece's effectiveness.Meanwhile a tremor time ratio was brought into the method to dispose the discontinuous tremor.Finally,the recognition method is verified with videos and records,and the sensitivity is 91.7% and specificity is 90.9%.Due to the FOG's contingency,in this paper a walking-cognitive double task experiment and a single task walking contrast experiment are designed to induce FOG.The recognition method is designed according to the 2~6 hz tremor of the leg when FOG occurs.The time domain gait cycle recognition is introduced into this method to improve the traditional PSD FOG recognize method.Normal walk,FOG and turning gait markers are recognized based on spectrum,energy and time domain characteristics.Finally,in the FOG recognition experiment,the sensitivity is 100%,the specificity is 88.9%,in FOG segment recognition the sensitivity is 84.7%.This paper explored the early discovery and diagnosis of PD based on the previous research and other typical motor manifestations of PD including stiffness,bradykinesia and dyskinesias.The parameters are extracted and analyzed from the stability and asymmetry of patient's rapid alternating movements' cycle,integral,and so on,the comparison of single task and dual task walking gait parameters' stability and asymmetry,the displacement of palming and tremor in palming.Combined with the results of the tremor recognition and the FOG recognition results from the previous researches,the PD patients in the database were given a report of PD recognition,and the resulting sensitivity is 75% for the early PD patients and 72.53% for all PD,and specificity is 81.82%.