The application of feature selection techniques analysis for human electroencephalograms in epilepsy
Valentin A. Yunusov,1,2 Sergey A. Demin,1 Alexander V. Minkin,3
1. Institute of Physics, Kazan Federal University, Kazan, Russia
2. Institute of Computational Mathematics and Information Technologies, Kazan Federal University, Kazan, Russia
3. Yelabuga Institute, Kazan Federal University, Yelabuga, Russia
Abstract
Epilepsy is a neurological disease that leads to the manifestation of seizure of its patients. The analysis of bioelectric brain signals recordings – electroencephalograms (EEG) is widely used in the field of neurology for diagnosis this pathology. In this study, we consider possible application of feature selection analysis implemented in the Weka software package for the diagnosis of epilepsy. We analyze bioelectric brain signals for solving the problem of classifying EEG signals obtained from two groups of people: patients with epilepsy and healthy subjects. The spontaneous nonparoxysmal or background EEG signals were recorded by placing scalp electrodes arranged in accordance with the 10–20 international electrode placement system. The subjects were instructed to close their eyes for some time during the recording. Features were extracted from EEG signals using Python framework eeglib, which is compatible with Weka package.
During the analysis, we considered 25 machine learning algorithms included in the Weka package, for which the classification accuracy exceeded 75% on the original feature space (over 100 features for classification). Analysis was performed in two stages. At the first stage, we analyzed signals recordings under study using two evaluators: CfsSubsetEval and WrapperSubsetEval within the initial feature space. After that, we chose new subset of features of each evaluator based on individual predictive power of features. Then, we trained classifying algorithms on obtained subsets of features for evaluating their effectiveness for solving classification problem. As a result of implementation of two-staged analysis, the highest classification accuracy increased from approximately 77% to over 85%. Best result was obtained for MultiLayerPerceptron algorithm for CfsSubsetEval evaluator. We believe that our study, after appropriate validation, will be helpful for the development and application of the new diagnostic methods for the patients with epilepsy.
Speaker
Valentin A. Yunusov
1. Institute of Physics, Kazan Federal University, Kazan, Russia; 2. Institute of Computational Mathematics and Information Technologies, Kazan Federal University, Kazan, Russia
Russia
Discussion
Ask question
