| Revista: | Computación y sistemas |
| Base de datos: | |
| Número de sistema: | 000560603 |
| ISSN: | 1405-5546 |
| Autores: | Sahu, Rekha1 Dash, Satya Ranjan2 Cacha, Lleuvelyn A3 Poznanski, R. R4 Parida, Shantipriya5 |
| Instituciones: | 1KIIT University, School of Computer Engineering, Orissa. India 2KIIT University, School of Computer Application, Orissa. India 3Sultan Zainal Abidin University, Faculty of Health Science, Terengganu. Malasia 4Sultan Zainal Abidin University, Faculty of Informatics and Computing, Terengganu. Malasia 5Idiap Research Institute, Martigny. Suiza |
| Año: | 2021 |
| Periodo: | Jul-Sep |
| Volumen: | 25 |
| Número: | 3 |
| Paginación: | 493-514 |
| País: | México |
| Idioma: | Inglés |
| Tipo de documento: | Artículo |
| Resumen en inglés | The mental illness or abnormal brain is recorded with EEG, and it records corollary discharge, which helps to identify the schizophrenia spontaneous situation of a patient. The recordings are in a time interval that shows the brain’s different nodes normal and abnormal activities. The spiking neural network procedure can be applied here to detect the abnormalities of patients. The abnormal spikes are detected using the temporal contrast method, and Poisson probability has been used to find the probability of abnormality discharge of each channel. Then recurrent neural network advance version long short-term memory trained with nine channels of probability values to generate the probability of spontaneous EEG activity during schizophrenia. On learning of a long short-term memory trainer, Adam gradient optimization technique is implemented. Finally, using decoded temporal contrast method schizophrenia patients predicted by the above procedure accuracy using cross-validation method predicted as 97% whereas actual positive rate showing computes the area under the receiver operating characteristic curve as 100% area. Again, after a threshold implement of the temporal contrast method, it is predicted 100% accuracy with the testing dataset. The novelty and robotic of a spiking neural network model called probabilistic spiking neuron model are shown after the mathematical formulation of input data set to generate the spikes carefully and intelligently like Hz value of EEG should be fixed accurately for the schizophrenia patients and selection of suitable recurrent supervised classifier. |
| Disciplinas: | Ciencias de la computación |
| Palabras clave: | Procesamiento de datos |
| Keyword: | EEG, Spiking neural network, Long short-term memory, Temporal contrast, Poisson probability distribution, Schizophrenia, Probabilistic spiking neuron model, Electroencephalography spikes, Data processing |
| Texto completo: | Texto completo (Ver HTML) Texto completo (Ver PDF) |
