Open AccessSimple Detection of Epilepsy From EEG Signal Using Local Binary Pattern Transition Histogram
Author(s) -
Muhammad Yazid,
Fahmi Fahmi,
Erwin Sutanto,
Wervyan Shalannanda,
Ruhush Shoalihin,
Gwo-Jiun Horng,
Aripriharta
Publication year - 2021
Publication title -
ieee access
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.587
H-Index - 127
ISSN - 2169-3536
DOI - 10.1109/access.2021.3126065
Subject(s) - aerospace , bioengineering , communication, networking and broadcast technologies , components, circuits, devices and systems , computing and processing , engineered materials, dielectrics and plasmas , engineering profession , fields, waves and electromagnetics , general topics for engineers , geoscience , nuclear engineering , photonics and electrooptics , power, energy and industry applications , robotics and control systems , signal processing and analysis , transportation
This paper proposed a simple but highly accurate feature extraction method for epilepsy detection from electroencephalogram (EEG) signals. Based on the combination of Discrete Wavelet Transform (DWT) and the newly proposed features Local Binary Pattern Transition Histogram (LBPTH) and Local Binary Pattern Mean Absolute Deviation (LBPMAD), our proposed feature extraction method can efficiently extract features from EEG signals for machine learning classification of epilepsy, achieving high classification accuracy with a feature size of only 18 for each signal. Tested on the publicly available University of Bonn Epilepsy EEG Dataset using a signal length of 4097 data points (23.61 seconds), the proposed method achieved larger than 99.6% accuracy results for Support Vector Machine (SVM) and k-Nearest Neighbor (KNN) classification of ictal (set E) against any non-ictal (set A, B, C, or D) or combinations of non-ictal (set $\text{A}+\text{B}$ , set $\text{C}+\text{D}$ , or set $\text{A}+\text{B}+\text{C}+\text{D}$ ) EEG signals, which is among the best of currently published works. Our method can maintain high classification accuracy even with short input signals, achieving more than 99.1% SVM classification accuracy when input signal length is reduced to 512 data points (2.95 seconds). The high accuracy, small feature size, ability to work with short input signals and low computing requirements made the proposed method suitable for mobile, low power, and low-cost wearable medical devices.