Article

Epileptic Seizure Detection Using Probability Distribution Based On Equal Frequency Discretization

Department of Electronics and Computer, Gaziosmanpasa University, Tokat, Turkey.
Journal of Medical Systems (Impact Factor: 1.37). 03/2011; 36(4):2219-24. DOI: 10.1007/s10916-011-9689-y
Source: PubMed

ABSTRACT In this study, we offered a new feature extraction approach called probability distribution based on equal frequency discretization (EFD) to be used in the detection of epileptic seizure from electroencephalogram (EEG) signals. Here, after EEG signals were discretized by using EFD method, the probability densities of the signals were computed according to the number of data points in each interval. Two different probability density functions were defined by means of the polynomial curve fitting for the subjects without epileptic seizure and the subjects with epileptic seizure, and then when using the mean square error criterion for these two functions, the success of epileptic seizure detection was 96.72%. In addition, when the probability densities of EEG segments were used as the inputs of a multilayer perceptron neural network (MLPNN) model, the success of epileptic seizure detection was 99.23%. This results show that non-linear classifiers can easily detect the epileptic seizures from EEG signals by means of probability distribution based on EFD.

0 Followers
 · 
179 Views
  • [Show abstract] [Hide abstract]
    ABSTRACT: This study proposes a new method, equal frequency in amplitude and equal width in time (EFiA-EWiT) discretization, to discriminate between congestive heart failure (CHF) and normal sinus rhythm (NSR) patterns in ECG signals. The ECG unit pattern concept was introduced to represent the standard RR interval, and our method extracted certain features from the unit patterns to classify by a primitive classifier. The proposed method was tested on two classification experiments by using ECG records in Physiobank databases and the results were compared to those from several previous studies. In the first experiment, an off-line classification was performed with unit patterns selected from long ECG segments. The method was also used to detect CHF by real-time ECG waveform analysis. In addition to demonstrating the success of the proposed method, the results showed that some unit patterns in a long ECG segment from a heart patient were more suggestive of disease than the others. These results indicate that the proposed approach merits additional research.
    Computers in biology and medicine 10/2013; 43(10):1556-62. DOI:10.1016/j.compbiomed.2013.07.015 · 1.48 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we used the Hilbert-Huang transform (HHT) analysis method to examine the time-frequency characteristics of spike waves for detecting epilepsy symptoms. We obtained a sample of spike waves and nonspike waves for HHT decomposition by using numerous intrinsic mode functions (IMFs) of the Hilbert transform (HT) to determine the instantaneous, marginal, and Hilbert energy spectra. The Pearson correlation coefficients of the IMFs, and energy-IMF distributions for the electroencephalogram (EEG) signal without spike waves, Spike I, Spike II and Spike III sample waves were determined. The analysis results showed that the ratios of the referred wave and Spike III wave to the referred total energy for IMF1, IMF2, and the residual function exceeded 10 %. Furthermore, the energy ratios for IMF1, IMF2, IMF3 and the residual function of Spike I, Spike II to their total energy exceeded 10 %. The Pearson correlation coefficients of the IMF3 of the EEG signal without spike waves and Spike I wave, EEG signal without spike waves and Spike II wave, EEG signal without spike waves and Spike III wave, Spike I and II waves, Spike I and III waves, and Spike II and III waves were 0.002, 0.06, 0.01, 0.17, 0.03, and 0.3, respectively. The energy ratios of IMF3 in the δ band to its referred total energy for the EEG signal without spike waves, and of the Spike I, II, and III waves were 4.72, 6.75, 5.41, and 5.55 %, respectively. The weighted average frequency of the IMF1, IMF2, and IMF3 of the EEG signal without spike waves was lower than that of the IMF1, IMF2, and IMF3 of the spike waves, respectively. The weighted average magnitude of the IMF3, IMF4, and IMF5 of the EEG signal without spike waves was lower than that of the IMF1, IMF2, and IMF3 of spike waves, respectively.
    Journal of Medical Systems 01/2015; 39(1):170. DOI:10.1007/s10916-014-0170-6 · 1.37 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: In Intelligent Tutoring Systems, affect-based computing is an important research area. Common approaches to deal with the affective state identification are based on input data from external sensors such as eye-tracker and EEG, as well as methods based on mining of ITS log data. Sensor based methods are viable in laboratory settings but they are tough to implement in real-world scenario which might cater to a large number of students. In our research, we create a mathematical model of frustration based on its theoretical definition. We identify the variables in the model by applying the theoretical definition of frustration to the ITS log data. This approach is different from existing data mining techniques, which use correlation analysis with labeled data. We apply our model to Mindspark, a commercial maths Intelligent Tutoring System, used by several thousand students. We validate our model with human observations of frustration.
    Advanced Learning Technologies (ICALT), 2012 IEEE 12th International Conference on; 01/2012