The Research On BECT Spike Detection Algorithm Based On Optimal Template Matching And Feature Fusion

Epilepsy is one of the chronic diseases of the brain caused by abnormal neuronal discharge,which is characterized by repeated and unpredictable seizures.More than 65 million people worldwide suffer from the dual physical and psychological effects of epilepsy,which is more common in infants and older people.Benign childhood epilepsy with centro-temporal spikes(BECT)is one of the most common childhood epilepsy,accounting for about 15%-20% of children with epilepsy.The prominent feature is the abnormal discharge of spikes near the central and temporal regions.Electroencephalogram(EEG)is a test tool that records the electrical activity of neurons.It is the most commonly used evaluation method to diagnose brain diseases and analyze the effect of drugs.Quantitative analysis of spikes is one of the most effective ways to help neurologists make diagnoses and provide therapeutic schedule.Traditionally,template-matching algorithms have been used by many researchers for their ability to extract spike-like waves on long-range EEG.However,the morphology of EEG spikes collected from different patients or the same patient at different times is very different,so the traditional template matching method is difficult to adapt to the EEG spikes of different patterns.In addition,current spikes detection algorithms tend to focus on the morphological characteristics of spike,but ignore the feature of spikes highlighting background activity.To solve these problems,a BECT spike detection algorithm based on optimal template matching and feature fusion is proposed.The main research work of this paper is divided into the following aspects:1.A BECT spike detection algorithm based on optimal template matching and morphological feature extraction is proposed.In this study,the sensitivity of the algorithm to spikes was improved by optimizing the universal template.First,the traditional universal template was used to match the spikes in EEG,and the result was clustered for spike detection.Then,the particle swarm optimization(PSO)algorithm was used to optimize the template parameters.Finally,false positive spike elimination based on bipolar(BP)channels and average reference(AV)channels could be implemented to achieve BECT spike automatic detection by taking advantage of the peak-to-peak characteristics of spikes discharge on BP channels and the accuracy of extracting AV channels spike patterns.2.A BECT spike detection algorithm based on background features and morphological features is proposed.In this study,background features were introduced to improve the performance of the spike detection algorithm.Firstly,spike-like extraction was used to extract spike-like from long-range EEG.The spike-like extraction was divided into two parts: optimal template matching and spike clustering detection.The dense needle type false positive spikes generated by electromyographic interference were eliminated,then the multi-stage background signal features of the remaining spikes were extracted,and the weights of these background signals were optimized using the harris hawks optimization(HHO)algorithm.Finally,the weighted background features were extracted using the obtained weights.Spikes morphological features and weighted background features are fused into feature vectors to train random forest(RF)model and complete BECT spike automatic detection.