Research Article
Classification Model of Spikes Morphology Using Principal Components Analysis in Drug-Resistant Epilepsy
@INPROCEEDINGS{10.1007/978-3-319-72965-7_27, author={Ousmane Khouma and Mamadou Ndiaye and Idy Diop and Samba Diaw and Abdou Diop and Sidi Farsi and Birahime Diouf and Khaly Tall and Jean Montois}, title={Classification Model of Spikes Morphology Using Principal Components Analysis in Drug-Resistant Epilepsy}, proceedings={Innovation and Interdisciplinary Solutions for Underserved Areas. First International Conference, InterSol 2017 and Sixth Collogue National sur la Recherche en Informatique et ses Applications, CNRIA 2017, Dakar, Senegal, April 11--12, 2017, Proceedings}, proceedings_a={INTERSOL \& CNRIA}, year={2018}, month={2}, keywords={Epilepsy Spike detection SNEO PCA Unsupervised classification}, doi={10.1007/978-3-319-72965-7_27} }- Ousmane Khouma
Mamadou Ndiaye
Idy Diop
Samba Diaw
Abdou Diop
Sidi Farsi
Birahime Diouf
Khaly Tall
Jean Montois
Year: 2018
Classification Model of Spikes Morphology Using Principal Components Analysis in Drug-Resistant Epilepsy
INTERSOL & CNRIA
Springer
DOI: 10.1007/978-3-319-72965-7_27
Abstract
Epilepsy is one of the diseases that are more subject to consultation in neurological clinics. To help neurologists to accurately diagnose this disease, several technological tools have been developed. Electroencephalography (EEG) of scalp or deep is a signal acquisition tool from electrical discharges of the brain areas. These signals are often accompanied by transient events commonly called interictal paroxystic events (IPE) or spikes of short durations. Analysis of these IPE could help with the diagnosis of drug-resistant epilepsy. With this intention, we will first of all seek to detect IPE, by separating them from the basic activity of signal EEG. In this paper, we propose spike detection method based on Smoothed Nonlinear Energy Operator (SNEO) using adaptive threshold. Then we will implement a new approach using principal components analysis (PCA) before classification to separate the events detected according to their morphologies. The objective in the long term is to characterize their space-time distribution over all the duration of the EEG signal.
