Research On Feature Extraction And De-redundant Compression Algorithm Based On High Frequency Oscillation Rhythm

With the development of economy,today's medical technology has been greatly improved compared with the past,but the detection of epilepsy is still a research hotspot.At present,the method of examining epilepsy is to find high oscillation rhythm by observing EEG signals through experienced doctors,and then judge the time of epilepsy seizures in most hospitals.This method has great limitations.The amount of EEG data is usually very large,so doctors need to spend a lot of time,which also affects the accuracy of judgment.Therefore,how to process EEG data efficiently is of great significance to doctors and patients.Based on the feature extraction and location of epileptic lesions,a de-redundancy compression algorithm based on prior template matching search is designed in this thesis.By using a method similar to traditional data compression,the efficiency of prediction and diagnosis of epileptic seizures in clinicians is greatly improved.The research content of this thesis mainly includes the following parts:Firstly,in terms of the data,the EEG signal is preprocessed and the signal is normalized to the interval [-1,1].Since the EEG signal is affected by the artifacts such as Eye Electricity,this thesis uses independent component analysis algorithm to remove artifacts,and then the band except 80-500 Hz is filtered out by using a Chebyshev type II IIR bandpass filter,and 50 Hz power frequency interference is removed by a comb filter to get the final EEG data.Secondly,in terms of the localization of epilepsy lesions,the threshold is set to screen out the suspected epilepsy initiation area,and the irrelevant data is eliminated due to the large amount of EEG signal data.The algorithm selected in this thesis is power spectral density and fuzzy entropy,and then the data of the suspected epilepsy initiation area are finally localized by wavelet time-frequency diagram.Finally,on the interception of data segments containing high frequency oscillation rhythm,a de-redundancy compression algorithm is proposed.From the angle of waveform of high frequency oscillation rhythm,a direct compression method similar to traditional data compression is used to intercept the effective data segments containing high frequency oscillation rhythm by morphological matching search algorithm,and the initial long-range EEG data is compressed to about an hour.This thesis focuses on the EEG data processing after the final localization of epileptic lesions,Select the final positioning of the lead one interception prior template data segment,and optimizes the database of the prior template.At the same time,the template data segments are used to match and search to get data segments effective for the final diagnosis of clinicians.The simulation results also show that the proposed de-redundancy compression algorithm has high time compression ratio and effective data segment ratio,which can meet the clinical needs well.