A novel method for removal of in-band noise and accurate extraction of information from ECGs

Abstract

Listen

Translate article

High pass filter Many publications have documented the issues associated with noise in ECGs collected in safety pharmacology studies. The presence of noise can result in excess variability in interval measurements, make arrhythmias difficult to detect, and confound ECG analysis software to the point where a large percentage of cardiac cycles may be uninterpretable. The affects of noise on ECG analysis can also increase the sample size required to detect QT prolongation. To address these issues, techniques have been developed that target reducing noise levels in acquired ECGs. Most problematic is noise that has energy in the same part of the frequency spectrum occupied by the ECG signal. This is generally caused by EMG and by changes in the tissueelectrode interface. The most common technique used to remove this noise is band-pass filtering (BPF). However, BPF is problematic in that using it to reduce in-band noise will result in distortion of the ECG. Multi-Domain Signal Processing (MDSP) is a novel approach that removes in-band noise without distorting the ECG signal. It works by independently isolating noise and other aspects of the signal in multiple mathematical domains. Once noise is isolated, it can be removed without distorting the ECG. MDSP also leverages the ability to isolate features of the ECG to provide for accurate interval measurement, identification of arrhythmias, and shows promise for measurement of respiratory rate by isolating aspects of the ECG that change during the respiratory cycle.