ProcessSignals. Detection of real EGG and BVP Peaks from noisy biosignals: an innovative MATLAB-based Graphical User Interface (GUI)

Deepak Panday,1* David Mayor,2 Hari Kala Kandel3

Abstract for a poster presentation, GUI). 2nd Annual Engineering and Computer Science Research Conference, University of Hertfordshire, 8 April 2020 (postponed).

Context. The variability of inter-peak intervals in electrocardiogram (ECG) or blood volume pulse (BVP) signals is a key measure of autonomic (parasympathetic) health and a frequent research topic. Pulse transit time (PTT) between ECG ‘R’ peak and the ‘foot’ of the following BVP peak is a useful indicator of arterial stiffness and ECG T-wave amplitude may enable assessment of autonomic (sympathetic) activity. In our previous work, we found that electrical stimulation to the hands may have frequency-specific effects on heart rate variability and blood flow. The unique ECG-BVP graphical user interface (GUI) described here (Fig. 1) was developed to help determine whether PTT and ECG T-wave amplitude confirm or refute our previous findings.

Methods & Objectives. The GUI was designed to (i) facilitate selection of different filters with options to change parameters values, (ii) identify ECG Q, R, S and T waves (peaks), and feet, peaks and ends of the BVP waveform, (iii) detect false (noise) peaks and (iv) facilitate adding, deleting or relocating ECG or BVP peaks. Firstly, we denoised the signals using a median filter to remove spikes and low- and high-pass IIR filters. Baseline (low frequency) wander was then removed and finally we detected any noise-induced conflicts between false ECG and/or BVP peaks and removed them manually. The strong correlation between locations and numbers of ECG (R, T) and BVP peaks could be used to identify appropriate filter parameters and removal of false peaks. The GUI has been tested initially on 1993 5-minute biosignal recordings sampled at 1024 Hz and collected using a NeXus10 amplifier with BioTrace+ software in the Physiotherapy Lab at the University of Hertfordshire in 2015-16.

Results. This novel GUI performed well, and results are currently being analysed. In future work, we plan to design a decision tree-based algorithm where the whole process discussed above can be automated.

[Link to software]

1. School of Engineering and Computer Science, University of Hertfordshire; 2. School of Health and Social Work, University of Hertfordshire; 3. Hertfordshire International College; * Corresponding author