| Title |
Elektroencefalogramų analizės metodų tyrimas / |
| Translation of Title |
Investigation of automatic EEG analysis algorithms. |
| Authors |
Misiukas Misiūnas, Andrius Vytautas |
| Full Text |
|
| Pages |
116 |
| Keywords [eng] |
EEG ; classification ; machine learning ; artificial intelligence |
| Abstract [eng] |
Automatic algorithm for electroencephalogram (EEG) classification by diagnosis: benign childhood epilepsy with centrotemporal spikes (rolandic epilepsy) (Group I) and structural focal epilepsy (Group II) are presented in this thesis. Manual classification of these groups is sometimes difficult, especially when no clinical record is available, thus presenting the need for an algorithm for automatic classification. A few possible classification by diagnosis algorithm versions are proposed in this thesis: 1) geometric EEG spike parameter and feed-forward multilayer perceptron (MLP) based classifier achieving 75% classification accuracy; 2) extremely randomized tree based algorithm using signal in channel where EEG spikes are classifying 82% accuracy; and 3) convolutional neural network (CNN) and majority rule classifier based algorithm achieving 80% accuracy, or 82% if only EEGs with 100 or more spikes are classified. |
| Dissertation Institution |
Vilniaus universitetas. |
| Type |
Doctoral thesis |
| Language |
Lithuanian |
| Publication date |
2020 |
