Study On Early Diagnosis Of Parkinson’s Disease Based On Electroencephalogram(EEG)

Parkinson’s disease(PD),also known as tremor paralysis,is the third most common disease among the elderly.Tremor,myotonia and decreased movement are its main clinical features.The early symptoms of Parkinson’s disease are not obvious,and patients are often in the middle and late stages of Parkinson’s disease when various clinical indicators are observed.In order not to miss the best opportunity for early treatment,the early diagnosis of Parkinson’s disease is of great importance.In recent years,in addition to changes in motor symptoms in patients with Parkinson’s disease,researchers are increasingly concerned about other non-motor symptoms.Sleep disorders have been recognized as one of the most common symptoms of Parkinson’s disease,and can precede clinical motor symptoms.Research shows that in the early stage of Parkinson’s disease,partial electroencephalogram(EEG)activity of patients has changed.Therefore,this paper aims to find out the characteristics of sleep EEG in patients with Parkinson’s disease in the early stage based on the above phenomenon.During the sleep state,there is no recognized characteristic index of EEG in patients with Parkinson’s disease,therefore,this paper first designed a personalized experiment to obtain sleep EEG data of mice before and after suffering from Parkinson’s disease.Secondly,the wavelet scattering network(Scatnet)is adopted for the nonlinear,multi-scale analysis of the time and frequency domain of the sleep EEG signal,so as to explore the differences of the covariance structure and global dependence between the sleep EEG of Parkinson’s disease and the normal(before the disease)sleep EEG.Then,EEG signal,as a kind of typical time series signal,has certain causality and correlation between the time series.To better exploit its information between the sequence of these features,long short-term memory(LSTM)recursive neural network which has good performance in this aspect is adopted to learn the implicit characteristic differences between EEG of Parkinson’s disease and normal EEG that include basic frequency segment after wavelet decomposition and reconstruction.Finally,since the brain electrical signal is a complex signal which is nonstationary and nonlinear,and patients with Parkinson’s disease have the characteristic of slow brain activity,while instantaneous frequency(IF)describes the time varying regularity of frequency and power spectrum entropy(PSE)can not only reflect the spectrum structure features of brain electrical signal,and can be used as an index of complexity of nonlinear EEG signal.Therefore,this paper proposes a method to combine the feature extraction of instantaneous frequency and power spectrum entropy with LSTM neural network model,so as to explore the differences between normal EEG and Parkinson’s EEG in features of nonstationarity,nonlinearity,time-frequency domain and complexity.In the experimental part of this paper,the differences between EEG of Parkinson’s disease and normal EEG in different characteristics of mice in sleep stage are studied by using the methods mentioned above.F1 value and classification accuracy are used to analyze the above experimental results.The experimental results show that,the method of combining instantaneous frequency and power spectrum entropy characteristics with LSTM neural network model which proposed in this paper has achieved the best effect,which can distinguish the EEG of Parkinson’s disease from normal EEG,thus laying a foundation for the early diagnosis of Parkinson’s disease.