The software program AMSIMP is used to score the locations of the B- and X-point, necessary to compute PEP and LVET. The number of algorithms proposed to score the impedance cardiogram is countless. We have tried quite a few at the Vrije Universiteit. Our current stance is that automatic scoring simply will not work. We therefore score every ensemble average by hand. This approach has reached quite reliable results so far (see literature). As a matter of fact, it leads to high correspondence with laboratory based signals that were scored with the Nihon Kohden IC equipment and with complete ECG and phonocardiogram traces available (Willemsen et al., Psychophysiology, 1996).
One waveform (complex) on the screen represents (depending on your previous settings in AMScom) a minute mean of impedance changes due to blood flow. If the file is labeled (ADVISORY, almost OBLIGATORY if you have prolonged ambulatory measurements) you can use the ensemble averages, which is a label average of the one-minute means. In this ensemble average (which is a valid way of handling large amounts of impedance data, see Riese et al., 2003) three points need to be identified:
- B-point or upstroke, the opening of the aortic valves, marking the end of the electromechanical systole and the beginning of the left ventricular ejection time. The upstroke occurs somewhere in the middle of the first heart sound. An increase in heart rate should be accompanied by a shift of the B-point to the left (increased sympathetic activation) or no shift at all (parasympathetic withdrawal). A decrease in heart rate should be accompanied by a shift of the B-point to the right (sympathetic withdrawal) or no shift at all (increased parasympathetic activation).
- dZ/dtmin, the point where the velocity of ejection is at its maximum (in the graph, it is drawn in reverse polarity, so the minimum shows as a maximum in AMSIMP)
- X-point or incisura, the closing of the aortic valves, marking the end of left ventricular ejection time. The X-point corresponds well to the first high frequency component of the second heart sound. An increase in heart rate should be accompanied by a shift of the X-point to the left (either by increased sympathetic activation, or parasympathetic withdrawal) or no shift at all. A decrease in heart rate should be accompanied by a shift of the X-point to the right (increased parasympathetic activation or sympathetic withdrawal) or no shift at all.
As a consequence of activity or stress, the morphology of the complex may change.
Below we have stated 5 scoring principles that are to be used during visual inspection and correction of the dZ/dt signal. These principles are shown in order of importance:
The B-point or upstroke should be at a first or second order zero-crossing in the dZ/dt signal. It should be close to the dZ/dt=0 line, and be the starting point of the longest uphill slope before the dZ/dtmin point. However, rather than appearing as a clear incisura, the B-point may sometimes take the form of a subtle inflexion and may vary considerably from beat to beat. It is therefore very important to inspect the dZ/dt signal closely in order to identify it. Occasionally, there is simply no clearly identifiable point that can be chosen to fit the above description of the B-point. In that case the point of the dZ/dt=0 crossing may be appropriate (see also Sherwood et al., 1990).
The dZ/dtmin is the highest point in the window between the B- and the X-point. It is possible that the dZ/dt signal shows a double maximum, a bit like rabbit ears. Then choose the highest of the two.
The X-point or incisura is always a local minimum after the dZ/dtmin. Often it is the lowest point in the entire signal, but not necessarily. In the ideal situation it can be seen as a sharp notch. It may be that two or more waveforms lie in close proximity of the lowest point in the dZ/dt signal. In some people there seems to be complete absence of a recognizable X-point. In these instances, it is difficult to distinguish the appropriate point associated with the aortic valve closure (see also Sherwood et al., 1990). A rule to keep in these more difficult situations is to search for a "W" in the waveform, then choose the second minimum (mostly, this minimum is followed by the longest uphill slope).This is the most clearly identifiable point for the X-point. However, strictly,(to make it even more difficult) the X-point lies a bit closer to the dZ/dt maximum, but without heart sounds present, this is not a point that can be reliably identified in majority of complexes to be scored.
Whatever point you choose, choose that point consistently. If a "less-than-ideal" upstroke is present in all complexes, but an "ideal" upstroke is present in some, choose the less-than-ideal one in all complexes, even those featuring a more "ideal" upstroke. Before starting to score the ICG, try browsing through the entire signal first by keeping the ‘page down’ button pressed. You can then decide which points can be most consistently identified, and this holds for both for the B-point and the X-point.
-3- in dubio abstine
You may have quite a lot of one-minute ensemble averages. Sometimes 2 out of the 5 ensembles are ugly because of movement artefacts. Don't try to make the best of these 2 if you feel pretty confident about the other 3 ensembles. The 3 good ones will give a good estimate of the ICG parameters during that particular period. Simply reject the other two. In general: when in doubt, reject that complex altogether.
-4- physiological plausibility
If you have doubts on whether the dZ/dt signal is correct, or should be rejected, you might use the following values as an indication of where the B- and X-point should be in an ideal situation.
HR: 40-60 --> PEP: 100-140 --> LVET 300-450
HR: 60-80 --> PEP: 90-130 --> LVET 250-400
HR: 80-100 --> PEP: 80-120 --> LVET 250-350
HR: 100-120 --> PEP: 70-100 --> LVET 200-300
HR: 120+ --> PEP: < 80 --> LVET 180-300
NB. These values are based on adult recordings. If your signal shows B- and X-points outside of these ranges, this does not at all mean that your dZ/dt signals should be discarded. Above table is just an extra aid...
-5- inter rater reliability
Reliability increases if two raters score the same data set. Score it independently, compute interrater reliability and then discuss it. Because of the consistency principle, it is not favorable to take interrater averages. Instead, one of both should be convinced that the other was right.
Keep score of the quality of your own rating. Sometimes, scoring is difficult and doubtful, at other times you feel pretty sure. After scoring you might want to generate three parameters for "scoring-quality". Make separate judgements for B-point scoring, X-point scoring and general signal quality on a scale from 0 (yuk!) to 10 (excellent!). Later on, in SPSS, request to see the mean of all parameters as a function of your quality rating.
At the end of visual inspection the corrected impedance cardiogram data with the codes and labels from the label data file are written to the ICG report files (".irp"). In contrast to label data, where a single HR, MSSD, MOT etc. value is computed over a variable length period, the .irp files yields PEP, LVET etc. values on every single one-minute ensemble average that was stored. This is a lot of data and a rigorous control of the impedance cardiogram report files is called for. This can be done in SPSS quite easily. Click on ‘example files’ in the menu for an example of such a SPSS file (imped.sps).
The reliability of Stroke Volume (SV) computed by impedance cardiography remains a heavily debated issue. With ambulatory SV the problems are even worsened, because movement artefacts and the lack of a phonocardiogram do NOT increase SV reliability. Basal thorax impedance (Z0) and the average distance between the front electrodes (L0) are not a real problem, but incorrect scoring of LVET or the dZ/dtmin amplitude are.
In addition, spot electrodes pick up only half of the impedance measured by band electrodes, and without a factor 2 correction, the Kubicek formula yields supraphysiological SV’s. If you must use SV, do so only in a within subject design.
Andrew Sherwood, Robert Kelsey, William R. Lovallo, Lorenz J.P. van Doornen (1990) Methodological Guidelines for Impedance Cardiography Psychophysiology, 27 (1): 1-23.
Willemsen G, De Geus EJ, Klaver CH, Van Doornen LJ, Carroll D. (1996) Ambulatory monitoring of the impedance cardiogram Psychophysiology, 33 (2): 184-93.
Harriette Riese, Paul F.C. Groot, Mireille van den Berg, Nina H.M. Kupper,Ellis H.B. Magnee, Ellen J. Rohaan (2003) Large-scale ensemble averaging of ambulatory impedance cardiograms Behavior Research Methods, Instruments, & Computers, 35 (3): 467-477.