Automatic ECG-based seizure prediction VLSI system with pipelined support vector machine

Abstract:
This paper proposes a new VLSI seizure prediction system through integrating heart rate variability (HRV) analysis and machine learning technique. The prediction of epilepsy seizure based on the electrocardiogram (ECG) signal is very valuable because the ECG signal can be acquired much more easily than the electroencephalograph (EEG) signal. The presented VLSI prediction system consists of R-peak detection module, feature extraction module, SVM classification module. The R-peak detection is based on a simplified method, and the pipeline method is used in classification module to accelerate the prediction. We verify the presented system on a field-programmable gate array and the result shows that the system is fully functional and exhibits a prediction sensitivity of 86% and false positive rate of 1.5 times per hour.