Feature Extraction Method Of EEG High Frequency Oscillation Rhythm Based On Prior Template

Epilepsy is one of the more common neurological diseases,and it is a brain disease that affects people of all ages.The latest research shows that patients who cannot be successfully treated with anti-epileptic drugs require surgery to treat epilepsy.The key to epilepsy surgery is to accurately locate the epileptic seizure initiation area before surgery.High-frequency oscillation rhythm and epileptiform discharge can be used to locate the lesion,and the effect of high-frequency oscillation rhythm localization is better than that of epileptiform discharge.Therefore,in recent years,high-frequency oscillation rhythm has been widely used in epilepsy surgery and has become a localization index of preoperative epilepsy lesions.In this paper,on the basis of studying the feature extraction of high-frequency oscillation rhythm of epileptic EEG,two innovative extraction algorithms for high-frequency oscillation are proposed.The first is the method of feature extraction and location of epilepsy initiation area based on multi-feature fusion,which solves the problem that a single feature quantity cannot fully characterize the characteristics of epilepsy EEG;The second is an extraction method of high-frequency oscillation rhythm characteristics based on a priori template,which realizes accurate positioning of the epilepsy initiation area and simplifies the signal processing steps.These two algorithms improve the efficiency of clinicians in diagnosing epilepsy,and also provide doctors with some predictive information which improves the doctor’s predictive efficiency.First of all,this paper studies the preprocessing of high-frequency oscillation data.The preprocessing process includes normalization processing,IIR filter filtering,50 Hz multiplier power frequency notch filter filtering and blind source separation method to remove artifacts,etc.The pre-processing of data lays the foundation for the extraction and analysis of features below.Secondly,the most important step in EEG data processing is to extract features.In this paper,the characteristics of the high-frequency oscillation rhythm are extracted from the time domain and frequency domain.The main extracted features are power spectral density,short-term energy estimate and fuzzy entropy.According to these three characteristics,a method based on mult-feature extraction and positioning of epilepsy initiation area is proposed.The results of these three feature extraction methods are comprehensively analyzed,and combined with Support Vector Machine(SVM)classification technology.The false positive data is eliminated,and the lead suspected of epilepsy is screened out.Then,the final location is obtained through expert judgment.Finally,after the final lesion-causing lead is obtained according to the multivariate feature extraction and positioning the epilepsy initiation area algorithm,the lead data corresponding to this part of the feature is used as the prior template,and an algorithm for locating lesions based on prior template matching is proposed.By using the previously created template,two ways are used to match the data to be tested.One way is to use a waveform shape template to match the lead data to be detected,and the other way is to use a feature template to match the characteristic amount of the lead to be detected.If the template has a high correlation with the lead data to be detected,it means that the lead to be detected may be the initiation zone of a suspected epileptic seizure.The algorithm improves the efficiency of signal processing,and the data obtained can also be used to improve the efficiency of doctors’ diagnosis and prediction.