Feature Selection Using F-statistic Values for EEG Signal Analysis

Electroencephalography (EEG) is a highly complex and non-stationary signal that reflects the cortical electric activity. Feature selection and analysis of EEG for various purposes, such as epileptic seizure detection, are highly in demand. This paper presents an approach to enhance classification performance by selecting discriminative features from a combined feature set consisting of frequency domain and entropy based features. For each EEG channel, nine different features are extracted, including six sub-band spectral powers and three entropy values (sample, permutation and spectral entropy). Features are then ranked across all channels using F-statistic values and selected for SVM classification. Experimentation using CHB-MIT dataset shows that our method achieves average sensitivity, specificity and F-1 score of 92.63%, 99.72% and 91.21%, respectively.

[1]  J. Richman,et al.  Physiological time-series analysis using approximate entropy and sample entropy. , 2000, American journal of physiology. Heart and circulatory physiology.

[2]  He Chen,et al.  Characterizing dynamics of absence seizure EEG with spatial-temporal permutation entropy , 2018, Neurocomputing.

[3]  Keshab K. Parhi,et al.  Seizure detection using regression tree based feature selection and polynomial SVM classification , 2015, 2015 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC).

[4]  G. Flint,et al.  Seizures and epilepsy. , 1988, British journal of neurosurgery.

[5]  Claus Thorn Ekstrøm,et al.  Introduction to Statistical Data Analysis for the Life Sciences , 2014 .

[6]  Yao Ding,et al.  Automatic Epileptic Seizures Joint Detection Algorithm Based on Improved Multi-Domain Feature of cEEG and Spike Feature of aEEG , 2019, IEEE Access.

[7]  Ali H. Shoeb,et al.  Application of machine learning to epileptic seizure onset detection and treatment , 2009 .

[8]  Mehrdad Heydarzadeh,et al.  Automated EEG-Based Epileptic Seizure Detection Using Deep Neural Networks , 2017, 2017 IEEE International Conference on Healthcare Informatics (ICHI).

[9]  D. Ruppert.  The Elements of Statistical Learning: Data Mining, Inference, and Prediction , 2004 .

[10]  Pablo F. Diez,et al.  Patient non-specific algorithm for seizures detection in scalp EEG , 2016, Comput. Biol. Medicine.

[11]  Lili Chen,et al.  Automatic Diagnosis of Epileptic Seizure in Electroencephalography Signals Using Nonlinear Dynamics Features , 2019, IEEE Access.

[12]  Ling Huang,et al.  Feature Extraction of EEG Signals Using Power Spectral Entropy , 2008, 2008 International Conference on BioMedical Engineering and Informatics.

[13]  Markos G. Tsipouras.  Spectral information of EEG signals with respect to epilepsy classification , 2019, EURASIP J. Adv. Signal Process..

[14]  B. Pompe,et al.  Permutation entropy: a natural complexity measure for time series. , 2002, Physical review letters.

[15]  Junjie Chen,et al.  The detection of epileptic seizure signals based on fuzzy entropy , 2015, Journal of Neuroscience Methods.

[16]  Pietro Liò,et al.  A new approach for epileptic seizure detection: sample entropy based feature extraction and extreme learning machine , 2010 .

[17]  Javad Birjandtalab,et al.  Automated seizure detection using limited-channel EEG and non-linear dimension reduction , 2017, Comput. Biol. Medicine.