simple software for impedance cardiography
TRANSCRIPT
Behavior Research Methods, Instruments, & Computers1990, 22 (3), 317-318
Simple software for impedance cardiography
MIKE HARRIS, DOUGLAS CARROLL, and GWEN CROSSUniversity of Birmingham, Birmingham, England
Simple software techniques for the digital analysis of impedance cardiography data are described.The software digitizes simultaneous impedance and ECG signals, averages these signals overtime on a beat-by-beat basis, processes the averaged signals to locate relevant features, and computes heart rate, stroke volume, and cardiac output for the averaged period. The source and executable code, written in Pascal for a PDP-ll with standard peripheral addressing and a VT125graphics terminal, is available from the authors.
Impedance cardiography is a convenient, noninvasivetechnique for estimating cardiac output. However,although the recording apparatus is simple and cheap,commercial packages that include the necessary signalprocessing are generally expensive and often inflexible.For these reasons, several laboratories (see Cowings,Naifeh, & Thrasher, 1988; Sherwood, Allen, Obrist, &Langer, 1986; van Doomen & de Geus, 1989) have developed their own signal processing software to run onstandard laboratory mini- or microcomputers. This software can usually provide reliable estimates on a beat-bybeat basis; but such a level of precision requires considerable computing time and power, and adjustment may berequired for different experimental situations (e.g., restvs. exercise, which can produce dramatic changes in theshape of the impedance waveform). Here we report somevery simple techniques, which are easy to implement andquick to run, but which provide an adequate basis for ageneral, robust, quasiautomatic package for estimatingcardiac output on a minute-by-minute basis. Observationsconfirm that the results produced by these simple techniques are in very good agreement with those of other,more sophisticated, systems.
Waveform DigitizationWe begin by digitizing simultaneous recordings of ECG
and the first temporal derivative of impedance (Z'). Theseraw signals are provided by, for example, the MinnesotaImpedance Cardiograph ModeI304B. Digitization at 250samples/sec to 12-bit accuracy can be performed eitheron- or off-line, using a suitable FM tape recorder. Lowersampling rates would no doubt be adequate, given speedor storage limitations. It is common practice also to digitize a recording of heart sounds, but we have not foundthis particularly useful, especially when recording fromactive subjects.
This work was supported by MRC Grant G880577N. We gratefullyacknowledge Lorenz van Doomen and Eco de Geus of the Free University, Amsterdam, for their generous hospitality, advice, and practicalassistance. Correspondence should be addressed to Mike Harris, Schoolof Psychology, Cognitive Science Research Centre, The University ofBinningharn, Binningharn Bl5 2TI, England.
Detection of ECG R-Waves and Signal AveragingThe remainder of the analysis is performed off-line.
ECG R-waves are first detected and then used to synchronize the averaging of successive heartbeats in the ECGand impedance signals. R-waves are detected by findingthe first maximum following three consecutive intersampleintervals, each with a positive difference greater than 10%of the typical signal range (which is adjusted prior todigitization to give a I-V maximum positive deflection).This typically achieves better than 99.5% correct detection, even in quite noisy signals. However, since this stageunderpins all further processing, we generally check toseethat the standard deviation is less than 25% of the meaninterbeat interval in each 5-sec block. Blocks that fail thistest can be inspected visually and, if necessary, manually edited.
One-second segments of the ECG and impedance signals, starting 160 msec before each R-wave, arethen averaged over the required period. We typically average overl-min periods, since this provides sufficient resolution forexperimental purposes and achieves adequate noise reduction, even for signals from exercising subjects. The averaged heartbeat records can then be processed without anyadditional filtering.
Waveform AnalysisFigure 1 shows typical averaged ECG and impedance
signals, together with the features extracted from themby the procedures described here.
The ECG Q-wave is reliably detected as the last sample,s., of the monotonically decreasing 12-msec segment thatis closest in time before the R-wave (i.e., s, < St-l < St-2).
In practice, this is found by stepping backward fromthe R-wave.
The impedance peak (Z) is simply the point with thelargest value during the 400 msec following the ECGR-wave. The value Z:nax is the difference between the impedance at this point and at the B-point.
The impedance B-point (onset of left ventricular ejection) is more difficult to extract, because the notch visible in Figure 1 tends to disappear during exercise. Themost satisfactory and general criterion we have found isto use the first sample, St, on the ramp before the im-
317 Copyright 1990 Psychonomic Society, Inc.
318 HARRIS, CARROLL, AND CROSS
o
Q)
-g Q R....,'M~
~ IMPEDANCE
o
y t-I--'--------"'---------'---------'
o
B Z
250Time (ms)
X
500 750
Figure 1. Relevant features of typical ECG and impedance waveforms, produced by averaging heartbeats over a I-min period and using the proceduresdescribed in the text.
pedance peak, at which the slope increases by a factorof two or more (i.e., [S'+I-S,] ~ 2[s,-s,-d).
The impedance X-point (completion of left ventricularejection) is detected as the leading edge of the first monotonically increasing l2-msec segment during the 600 msecfollowing the impedance peak, providing that the valueat that point is within 10% of the minimum value duringthat period. On the very rare occasions when these criteria are not met, the X-point is left undefined.
Waveform Inspection and Estimation ofCardiac Performance
The values extracted for each segment can be visuallychecked and, if necessary, edited before further calculation. Since the data are reduced to l-min periods, this isa fairly fast process even for lengthy recordings. We havefound, in practice, that the estimates require very littleediting despite wide variations in recording techniques andsubject activity.
The final values are used to estimate pre-ejection period(PEP), as the time between the ECG Q-wave and theimpedance B-point, and left ventricular ejection time(LYET), as the time between the impedance B-point andthe X-point. These estimates can in turn be used to estimate, for example, stroke volume (SY) using the standard formula:
SV = eL~Z'max·LVETIZ~,
where e is 1350/cm, L o is the mean distance betweenthe recording electrodes, and Zo is the mean thoracic im-
pedance (provided by the cardiograph). Cardiac outputis simply the product of stroke volume and heart rate (obtained from the ECG R-waves).
The above procedures are currently implemented,mainly in Oregon Pascal-2, on a PDP-1l/34 minicomputer, running RTil, with a YTI25 graphics terminal.The AID converters used for digitization form part of aCED 502 laboratory interface, have a resolution of 12 bits,a total conversion time of 16 usee, and conform to standard DEC addressing and commands. The current implementation requires 256K RAM but, at some cost inspeed, it could easily be modified to run in 64K. Thesource and executable code may be obtained by contacting the authors via conventional or electronic mail (Email:[email protected]).
REFERENCES
COWINGS, P. S., NAiFEH, K., &0 THRASHER, C. (1988). A computer program for processing impedancecardiographic data: Improving accuracythrough user-interactive software (NASATechnical Memorandum 101020).
SHERWOOD, A., ALLEN, M. T., OBRlST, P. A., &0 LANGER, A. W.(1986). Evaluation of beta-adrenergic influenceson cardiovascularand metabolic adjustments to physical andpsychological stress. Psychophysiology, 23, 89-104.
VAN DooRNEN, L. J. P., &0 DE GEUS, E. J. C. (1989). Aerobic fitnessand the cardiovascular response to stress.Psychophysiology, 26, 17-28.
(Manuscript received January 18, 1990;revision accepted for publicationMarch 15, 1990.)