Classificação de sinais eletroencefalográficos utilizando Transformada Wavelet Discreta e Máquina de Vetores de Suporte: uma aplicação na diferenciação entre crises epilépticas e crises não epilépticas psicogênicas
Gespeichert in:
| 1. Verfasser: | |
|---|---|
| Publikationsdatum: | 2018 |
| Format: | Master thesis |
| Sprache: | por |
| Quelle: | Repositório Comum do Brasil - Deposita |
| Download full: | https://deposita.ibict.br/handle/deposita/41 |
Zusammenfassung: | The present work deals with the study and application of the Discrete Wavelet Transform (DWT) in conjunction with the Supporting Vector Machine (SVM) classifier in the differentiation between epileptic seizures and psychogenic non-epileptic seizures (PNES). A database with electroencephalogram (EEG) tests containing epileptic seizures and psychogenic non-epileptic seizures was collected at the Videoelectroencephalography Unit of the Institute of Psychiatry of the Hospital das Clínicas of the Medical School of the University of São Paulo (IPq-HCFMUSP). In the EEG signal processing, the Wavelet Discrete Transform (DWT) based on the Coiflet 1 and Daubechies 4 families and the direct signal extraction (without DWT) were used. From these processing, characteristic vectors were generated for the training and evaluation of the SVM classifier. In the analysis of the performance of the classifier, tests were performed by modifying the number of characteristics vectors for the classifier training, the origin of the characteristic vector (Coiflet 1, Daubechies 4 and direct extraction) and the kernel type (Linear, Polynomial , Radial Base Function - RBF - and Sigmoid). As a result, in the case of the use of 1-second windows in the EEG signal processing, the classifier was able to achieve a hit rate (accuracy) of up to 100% using the Linear kernel and the Coiflet 1 and Daubechies 4 families. In the case of the use of the total time of each crisis, the classifier obtained a hit rate of up to 100% in the four kernel types using the Coiflet 1 family. Thus, based on the feature vectors used, it was possible to conclude that the classifier SVM is efficient and its use is feasible in the differentiation between epileptic seizures and PNES. |