How to deal with ECG noise
Today heart diseases are one of the major causes of deaths worldwide and therefore it is necessary to have a proper method which determines the patient’s cardiac condition. Doctors use electrocardiograms (ECGs) to detect abnormal heart rhythms and to investigate the underlying chest pain sources.
An ECG signal consists of very low frequency signals from about 0.5 Hz – 100Hz: P wave, QRS complex and T wave; and any deviation in these parameters indicates the presence of cardiac abnormalities. A common problem in ECG analysis is the removal of unwanted artefacts and noise. Various artefacts get implemented and change the original signal, such as: electrode contact noise, power-line interference, muscle noise, etc. so if we want to avoid ECG misinterpretation, it is essential to reduce noise.
Persistent noises have similar temporal distributions in all leads, but with different intensity levels. The low-frequency range signifies the baseline wander, the medium frequency signifies the power-line interference and the high frequency signals signify the electromyography noise.
Electrode contact noise and motion artefacts
Variations in electrode-skin impedance and activities like patient’s movement and breathe cause baseline wander. Typically, it’s dominant on frequencies smaller than 0.5 Hz and represented by changing position of the isoelectric line.
The most simple and fastest way to remove it is to use linear time-invariant high-pass filter and cut off the lower frequency components. The cutoff frequency should be selected in a way that the ECG signal information remains undistorted while as much as possible of the baseline wander is removed. It depends on the slowest heart rate, which during the bradycardia can drop to 40 bpm, implying that the lowest cutoff frequency should be 0.67 Hz. However, a heart rate is not perfectly regular and it always fluctuates from one beat to another, so it is wiser to choose a slightly lower cutoff frequency such as 0.5 Hz.
The other crucial filter design consideration is the use of linear phase filter in order to prevent phase distortion from altering various wave properties of the cardiac cycle, such as the QRS complex duration, the ST–T segment level or the endpoint of the T wave. The FIR filters can have an exact linear phase response and usually provide better baseline removal, but their designs result in a high filter order (around 300). The other option is to use the forward-backward IIR filters. They meet the magnitude specification more easily with a much lower filter orders (2 or 3) and the use of forward-backward filtering ensures a zero-phase transfer function. Most authors recommend the use of FIR filter with the Kaiser window, but ultimately the choice of filter depends on the best fit to your data.
Therefore, to remove the baseline wander, the usual sampling rate alteration technique consists of:
- Decimation of the original signal which includes anti-aliasing filtering,
- Lowpass filtering to produce an estimate of the baseline wander,
- Interpolation of the estimate back to the original sampling rate and
- Subtraction of the estimate from the original ECG to produce the baseline-corrected signal.
Electromagnetic fields caused by a power line represent a common noise source in the ECG that is characterized by 50 or 60 Hz sinusoidal interference, possibly accompanied by a number of harmonics. Such narrow band noise cause problems interpreting low amplitude waves because it introduces unreliable and spurious waveforms. The traditional approach is to use a digital notch filter characterized by a unit gain at all frequencies except at the notch frequency where the gain is almost zero. The IIR notch filter with a narrow 3 dB rejection bandwidth is preferred to faithfully separate the sinusoidal and broadband components. Tolerable signal distortion needs a narrow frequency band, which leads to ineffective filtering with larger power-line frequency deviation. The linear time-invariant notch filters may introduce significant distortion in the QRS and ST-segment portions due to the filter ringing, thus the filters with a nonlinear structure could be a better choice.
EMG noise is caused by the contraction of other muscles besides the heart. When other muscles in the vicinity of the electrodes contract, they generate depolarization and repolarization waves that can also be picked up by the ECG. The extent of the crosstalk depends on the amount of muscular contraction (subject movement) and the quality of the probes. The EMG amplitude is stochastic and can be reasonably modelled by a Gaussian distribution function. The mean of the noise can be assumed to be zero and the standard deviation is typically 10% of the peak-to-peak ECG amplitude. EMG’s frequency spectra overlaps with ECG’s, so acceptable compromise wold be to use low-pass filter with the cut-off frequency of 40 Hz. This noise is common in subjects with uncontrollable tremor, disabled persons, kids and persons fearing the ECG procedure.
Burst noises are typically classified as a white Gaussian noise which appears on a subset of leads for a very short duration, such as: electrode pop noise, electrode motion artefact, electrosurgical noise, instrumentation noise etc. The frequency range of these noises isn’t well defined.
Electrode pop-up or movement noise
Variations in the position of the heart with respect to the electrodes and changes in the propagation medium between the heart and the electrodes initiate electrode contact noise. It is visible as sudden change in the amplitude of the ECG signal and low frequency baseline shift. Also, poor conductivity between the electrodes and the skin both reduces the signal amplitude of the ECG signal and increases the probability of disturbances. Sudden changes in the skin-electrode impedance induce sharp baseline transients which decay exponentially to the baseline value. This transition may occur only once or rapidly several times in succession. It is visible even after the removal of the baseline wander.
The electrical equipment used in ECG measurements also contributes noise. Electrode probes, cables, signal processor/amplifier, and the Analog-to-Digital converters are the major sources of this form of noise. This noise cannot be eliminated, but it can be reduced through higher quality equipment and careful circuit design.
So to sum up, there is no way to remove all the noises from all of your ECGs but you can design a reasonable filter. It should be highly accurate and as fast as possible.
- H. Limaye and V.V. Deshmukh (2016): ECG Noise Sources and Various Noise Removal Techniques: A Survey , IJAIEM, 5, 2, pp. 086-092 , ISSN 2319 – 4847.
- K.S. Kumar, B. Yazdanpanah, P.R. Kumar (2015): Removal of Noise from Electrocardiogram Using Digital FIR and IIR Filters with Various Methods, Communications and Signal Processing (ICCSP)