-
Notifications
You must be signed in to change notification settings - Fork 15
ECG Acquisition Hardware
This document is summarized from ECG Acquisition, Storage, Transmission, and Representation by Gari D. Clifford and Matt B. Oefinger.
Fluctuations in electrical potential between electrodes on the skin's surface will form a weak lead signal that must be amplified using an using an optically isolated instrumentation amplifier. In general, there will also be a third electrode for each differential signal to act as a ground. The lead signal is then passed through a highpass filter, a second amplification stage, then a lowpass filter, and finally sampled by an analogue-to-digital converter.
Any circuit that uses electrical power and comes in human contact must have the board designed into an isolated and nonisolated segment to prevent current leakage to the patient. The direct power source feeding the nonisolated board can be transferred to the isolated segment using DC-to-DC converters. Similarly, information is transferred from the patient (isolated) side of the board to the nonisolated side via opto-isolators. They are placed between the isolation gap and powered by the DC-to-DC converter output on the isolated side, and by the main power source on the nonisolated side. Note that high level resistors should be placed between each input signal (and ground) for static/defibrillation protection. Furthermore, a current limiting resistor at the output is required in case the op-amps fail. Also note that optical isolation in the early stage of amplification can introduce significant noise and thus often preferable to isolate directly after digitizing the signal.
Electrical systems within buildings, such as power-lines or mains, usually utilize AC power that run at frequencies around 60 Hz. Since these frequencies are well within the range of an ECG signal spectrum, it is often essential to format the acquisition circuit to remove any unwanted noise from other background electrical sources. This can be done using a notch filter in the 60 Hz range, but this will invariably remove some of the ECG signal content as well that could potentially yield vital information. Thus an active ground circuit is the preferred means of removing noise. An active ground circuit works at the preamplification stage by removing the common mode voltages from both electrodes of a lead, thus eliminating the electrical signals that are simultaneously present at both electrodes, allowing them to conduct only differential signals that stem from the heart.
High input impedance is imperative in biosignal acquisition as electrophysical signals of interest are often too weak to supply enough current to accurately detect and display. A CMOS circuit yields an extremely high input impedance and corresponding power amplification, which can serve as an ideal decoupling/buffer stage between the signal and analog signal processing circuit.
As mentioned in the Isolation and Protection section, introducing the isolation stage before amplification can produce significant noise to the signal. If subtleties in the ECG, such as late potentials, are not important, then it is possible to provide optical isolation at the preamplification stage. Isolation ensures that a power surge within the machine cannot affect the patient, and conversely prevents any surge coming from the inputs (such as defibrillation) cannot extend and damage beyond the preamplification stage. The preamplification stage must be able to account for differential voltages between electrodes of up to 100 mV, where a gain of 25 provides an adequate signal-to-noise ratio and does not saturate.
A cut-off of 0.5 Hz was widely used in traditional analog filtering, which reduced baseline drift due to the generally lower frequency of respiratory movement. However, this can lead to marked phase distortions in areas where frequency content and wave amplitude changes abruptly, particularly at the ST segment (after the QRS complex), leading to possible artefactual ST-segment deviation. Lowing the frequency cutoff to as low as 0.05 Hz as per the 1975 guidelines (and has since remained the current recommendation) will sequester the distortions of the repolarization waves but at the same time increase the level of baseline drift. Digital filtering allows for methods of increasing the low frequency cutoff (to avoid respiratory noise interference) without the introduction of phase distortion, such as using a flat step response filter that results in zero phase shift. The low frequency cutoff for linear digital filters with zero phase distortion is relaxed to 0.67 Hz (or below; this frequency corresponds to a heart rate of 40 bpm, well within the lower range of the vast majority of patients). The filter transfer function for the HP filter, which can also be used for the LP filter downstream, can utilize the Bessel transfer characteristic to minimize phase distortion. This optimization for phase response comes at the expense of a slow roll-off in the transition region.
After passing through the HP filter, the signal is amplified again, this time at a gain of 52. This second amplification stage further increases the SNR of the signal and boosts the signal voltage to a range appropriate for sampling with an A/D converter with a dynamic range of ± 10 V. The signal entering this amplification stage, in contrast with that entering the preamplification stage, is not offset due to half-cell potential differences and baseline drift because of the preceding HP filter stage. As such, this amplification stage can comfortably amplify the signal by the rather sizable factor of 52 without saturating the amplifiers.
High frequency is needed to accurately define the most rapidly changing parts of the ECG signal, namely the QRS complex, where detecting R amplitude and Q waves is essential diagnostic information. Data being sampled at 500 per second with a high frequency digital filter cutoff of 150 Hz is required to reduce amplitude error measurements ~ 1% in adults. Greater bandwidth of up to 250 Hz may be required for infants. Suboptimal frequencies such as 40 Hz that are used to reduce noise can potentially miss critical diagnostic information. As mentioned in the Highpass Filtering section, a filter with Bessel transfer characteristics will minimize phase distortion, but again at the expense of a slow roll-off in the transition region.
Oversampling is a technique often employed in systems using an antialiasing filter with relatively slow roll-off, such as the the highpass and lowpass filters aforementioned that can be used to minimize phase distortion. The sampling rate in this case may have to be higher than expected due to the slow roll-off. Please refer to pg.20 of http://www.mit.edu/~gari/ecgbook/ch2.pdf for a more detailed explanation.