Classifier Implementation for Spontaneous EEG Activity during Schizophrenic Psychosis

Abstract

The mental illness or abnormal brainis recorded with EEG, and it records corollarydischarge, which helps to identify schizophreniaspontaneous situation of a patient.The recordings arein a time interval that shows the brain’s differentnodes normal and abnormal activities.The spiking neuralnetwork procedure can be applied here to detect theabnormalities of patients. The abnormal spikes aredetected using temporal contrast method, and Poissonprobability has been used to find the probability ofabnormality discharge of each channel.Then recurrentneural network advance version long short-term memorytrained with nine channels of probability values togenerate the probability of spontaneous EEG activityduring schizophrenia. On learning of long short-termmemory trainer, Adam gradient optimization techniqueis implemented.Finally, using decoded temporalcontrast method schizophrenia patients predicted bythe above procedure accuracy using cross-validationmethod predicted as 97% whereas actual positiverate showing computes the area under the receiveroperating characteristic curve as 100% area.Again,after a threshold implement of the temporal contrastmethod, it is predicted 100% accuracy with the testingdataset.The novelty and robotic of a spiking neuralnetwork model called probabilistic spiking neuron modelare shown after the mathematical formulation of inputdata set to generate the spikes carefully and intelligentlylike Hz value of EEG should be fixed accurately forthe schizophrenia patients and selection of suitablerecurrent supervised classifier.