Electroencephalogram Analysis With Wavelet Transform And Neural Network As A Tool For Acute Ischemic Stroke Identification
The prevalence of stroke in Indonesia are 7% based on the health proffesionals statement and 12.1% based on patients' symptoms' history. Early examination using CT scan causes radiation effects and spent high operational cost whilethe prevalence of stroke based on diagnosis or symptoms were higher in the lowest (13.1 ‰) and mid lower (12.6 ‰) quintile of ownership index. This studytried to analyze the signals of EEG automatically based ontraining data sets from normal patients and patients with acute ischemic stroke (AIS) using digital signal processing such as wavelet transform and feedforward type of neural network with Extreme Learning Machine (ELM) algorithm. It was claimed that electroencephalography could help to confirm or detect acute ischemic strokewhich is shown by the presence of the slow wave and the asymmetrical wave of right and left hemisphere. This study uses Delta Alpha Ratio (DAR), Delta Theta Alpha Ratio (DTABR) and Brain Symmetry Index (BSI)'s value as the ELM input feature score which were obtained by Wavelet (Daubechies 4) transformation and Welch's method to identify acute ischemic stroke. In this study, the performance of systemtest accuracy, sensitivity and specificity were above 93%.
Keywords— Electroenchepalogram (EEG), Acute Ischemic Stroke, Wavelet Transform, Extreme Learning Machine (ELM).