WSEAS Transactions on Signal Processing
Print ISSN: 1790-5052, E-ISSN: 2224-3488
Volume 14, 2018
Identifying Outliers in HRV–Seizure Signals Using p-shift UFIR Baseline Estimates
Authors: , , ,
Abstract: Heart rate variability (HRV) is typically associated with neuroautonomic activity and viewed as a major non–invasive tool to detect seizures. The HRV has been assumed and analyzed as a stationary signal. However the presence of seizures can violate estimates of statistical parameters and conventional techniques intended to remove outliers can be inaccurate. A useful approach implies setting thresholds to compute the first and third quartiles from histogram data or residuals based on the estimated baseline. In this paper, we propose an accurate method to identify outliers in HRV measurements with partial epilepsy retaining relevant information. The baseline perturbed by the seizure in the HRV data is removed using the p-shift unbiased finite impulse response (UFIR) smoothing filter operating on optimal horizons. The residuals histogram is plotted and the upper bound (UB) and lower bound (LB) are computed as thresholds. A comparison is provided of a typical points detected in HRV/seizures based on several methods used to estimate the baseline. A time/frequency analysis is supplied to show the difference between the raw HRV and HRV without outliers. The method proposed is tested by partial seizures records taken from patients during continuous EEG/ECG and video monitoring.
Search Articles
Pages: 109-114
WSEAS Transactions on Signal Processing, ISSN / E-ISSN: 1790-5052 / 2224-3488, Volume 14, 2018, Art. #14