Design Of Epilepsy EEG Analysis System Based On Improved ELM

Epilepsy is a neurological brain disorder caused by excessively synchronized firing of neurons in the cerebral cortex.Currently,in clinical practice,the diagnosis of epilepsy is usually achieved by interpreting and analyzing the EEG recordings of patients.Aiming at the problem that the diagnosis of epilepsy requires long-term observation of the patient’s EEG,which not only increases the workload of doctors,but also affects the accuracy of analysis,the study proposes two improved extreme learning machine(ELM)algorithms.And this study completed the design of epilepsy EEG analysis system for EEG acquisition,processing and display and seizure detection.First,the study designs an EEG signal acquisition circuit with STM32F405RGT6 as the core and 24-bit ADS1299 as the analog-to-digital converter,which can acquire the subjects’ EEG signals in real time.And the research developed a host computer,which is used for filtering,real-time display and data storage of EEG signals,and at the same time,it uses EEG signal data to realize seizure detection and result display of epilepsy.Second,the study proposes a feature extraction method for epilepsy EEG signals,which combines three nonlinear dynamic features and sub-band significant statistical features after discrete wavelet transform to mine the effective information of epilepsy EEG signals.At the same time,considering the time-consuming problem of extracting multichannel EEG features,the study further uses convolutional neural networks to extract feature vector sets from EEG signals.Moreover,by embedding the particle swarm optimization algorithm into the genetic algorithm and improving the input weights and biases of the ELM,the PSO-GA-ELM epileptic seizure detection model was established,and it was used for the analysis of epilepsy EEG signals.Finally,this study uses the epilepsy EEG dataset to verify the algorithm model proposed in the paper,and compares it with the traditional ELM algorithm and another improved ELM algorithm,namely the WOA-MFO-ELM algorithm that optimizes the ELM by mixing the whale optimization algorithm and the moth-flame optimization algorithm.The study compares the PSO-GA-ELM model,WOA-MFO-ELM algorithm with the traditional ELM algorithm through the publicly available epilepsy EEG dataset.The results show that compared with the traditional ELM algorithm,the classification accuracy of the PSO-GA-ELM and WOA-MFO-ELM model are improved more than7.6% respectively,which proves their high classification accuracy and reliability.What’s more,the measured EEG data and the clinical epilepsy EEG data set provided by a provincial tertiary hospital were used for the comparative test of PSO-GA-ELM and WOA-MFO-ELM algorithms herein.The test results show that the classification accuracy of the two algorithms can reach more than 91.30%,and the average classification accuracy of the PSO-GA-ELM algorithm is 1.83% higher than that of the WOA-MFO-ELM algorithm.In conclusion,the improved ELM algorithm proposed here has important clinical application value for the analysis of epilepsy EEG signals.