As with all other tabs, the clock time and IBI time series are presented in the two bottom windows, to quickly locate our current position in the total recording. To further ease navigation, the labels (as labelled in the Label tab, section 5.2) are printed on top of the IBI time series.
Because the time frame of a single heartbeat is in the order of milliseconds, and complete recordings can be 24 hours or longer, two additional navigation bars are present, containing colored rectangles. Each colored rectangle corresponds to the signal segment drawn in the frame above it. By clicking, dragging, and resizing these rectangles you can navigate to the desired segment and resolution.
The ECG tab will assist you to create an artefact free IBI signal as quickly as possible by applying automated artefact and peak detection. Visual inspection and correction of the resulting IBI signal is facilitated by using automated detector for diverging beats, the QRS detector.
The QRS detector runs three separate automated analyses on the ECG signal. The first automated analysis detects and marks periods with missing data or clipping of the electrocardiographic signal. These periods are called artefacts. A second automated analysis of the QRST waveform detects the occurrence of all R-peaks. Each R-peak detected by the algorithm is marked in the top frame with a line (either blue, yellow or red).
The R-peaks are converted to the inter beat intervals time series which is simply the time in milliseconds between two consecutive R-peaks, and plotted as a continuous line in all navigation bars. The third automated analysis checks the plausibility of the duration of each IBI in the context of its surrounding inter-beat intervals.
TIP9: When your data is noisy or your participant has a non-standard ECG the default setting of the peak detection algorithm might need to be adjusted. This is addressed in Appendix A.
It is important to note any unexpected deviation in time between two consecutive beats is marked divergent by the algorithm. However, when a beat is marked red, this does not necessarily mean that the beat is invalid and something that needs to be corrected. This will be addressed further in section 5.3.3 Suspicious beat correction.
All actions for the ECG tab are presented in the form of buttons. The function of these buttons should be self-explanatory. Like in all other tabs hovering the mouse cursor over a button will display a short description. The last set of options is useful for assessing suspicious beats and to export beat-to-beat data.
Visual inspection and manual correction
Automated artefact labeling reliably detects clipping and signal loss, but detection of noisy ECG is not perfect. Manual selection of bad ECG signal parts may be additionally needed. The three main windows will help you select the parts of the IBI time series that need to be manually labeled as artefacts. The bottom window is our overview of the IBI signal of the entire recording that is also present in the other tabs of DAMS.
TIP10: view the “R-Peak Detection” tutorial video for a demonstration: www.vu-ams.nl/support/tutorials/software/r-peak-detection
In the main window with the ECG signal the detected R-peaks are marked by vertical lines, mostly blue. A blue line means that the beat was considered to be correct according to the automatic beat detector. Potential mistakes in automated beat detection are termed ‘diverging IBIs’ and are flagged by a red (deviant) or yellow (worth checking) color. The red marker beats are strongly divergent from the algorithm, while the yellow ones are only slightly divergent.
A useful strategy is to navigate through the most divergent beats to check and possibly adjust them until you reach several beats in a row which have been
detected correctly. You can navigate from most diverging to least diverging beat by pressing the period or > key on your keyboard. You can go backwards by using the comma or > key. You can also use the following buttons:
After your corrections (or in fact at any time) you can select the menu button re-check deviant IBI`s by clicking the circle with traffic light (or press ctrl + R) to see how many beats are still considered deviant. Just remember that deviant is not wrong per se!
A deviant beat can have several origins which will be discussed in the following order:
- Noisy data
- Misplaced R peaks
- Physiological artefacts
TIP11: In case of a very noisy long recording (24 hours for example), that would take too much time for the researcher to delete manually, VU-DAMS offers the option to mark all deviant beats as artefacts. You can find this feature by clicking the garbage can icon.
Below the ECG signal time axis is the bar named “ECG Artefacts”. All automatically detected artefacts by QRS detector software are labeled by a red bar. However, some parts of the data might not be bad enough to be detected by the automated artefact detector, while already too noisy for the R-peak detector to function properly. Hence the R-peak detector will try to make something out of noise and may still score occasional beats as being correct (blue) where they are not. These periods of noisy data will also contain a lot of red and yellow lines which makes them easy to detect.
The origin of the noise can be caused by either rigorous movements, detached electrodes or other technical difficulties. However, we handle this data in a similar manner: manually labeling them as artefacts. To mark the bad ECG as an artefact, click with the left mouse button in the artefact bar and drag it from left to right until it covers the entire period that you want to be marked as an artefact. When you let go of the mouse the following window will pop-up from which you can select the origin of the artefact (when unsure choose the option “Other”). All beats that fall within an artefact are deleted from all further data analyses. However, when <5 successive beats are deleted, the VU-DAMS software interpolates the IBI times using a cubic spline.
Example of a clipping signal caused by an electrode temporarily coming loose.
Misplaced r peaks
Example of noisy data due to movement of the torso:
TIP12: when removing large stretches of data containing artefacts, zoom out so the entire artefact period is visible in your screen.
When encountering a misplaced divergent beat several actions can be performed. You can delete a beat by right clicking on it, e.g. when a beat was placed in an obvious wrong location in between beats. Or you can add a beat by left clicking on the correct location of the R-peak, e.g. when a beat was completely missed. You can also move a beat by placing the left mouse button on the vertical line and then move it left or right. Releasing the mouse button will lock the vertical line to its new location. Notice that the surrounding beats might also change color when adding, deleting or moving a beat (see example below in which a yellow beat was moved slightly to the left).
In HRV analysis, the aim is to examine the sinus rhythm modulated by the autonomic nervous system. Hence, physiological artefacts should be excluded from analysis by marking them as artefacts. Physiological artefacts include ectopic beats and arrhythmic events or more generally: beats not originating from the sinus node. Normally, the electrical activity starts in the sinus node; the natural pacemaker of the heart. Electric current will travel through the atria and then is momentarily delayed in the AV node as a result of slow conduction. Thereafter, the current travels through the bundles of His to depolarize the ventricles in an organized manner (starting at the apex). This normal, organized electrical activity will result in the known P-QRS-T morphology.
Generally, there are two main types of ectopic beats; those originating from the ventricles and those originating somewhere else. An ectopic beat originating from the ventricles is called a premature ventricular contraction (PVC). You can recognize these beats by an early, broad QRS complex. Below, two examples of PVCs are shown (both are PVCs, the different morphology is due to a different site where the PVC originates in the ventricles). You can see the beat is early (premature) since, looking at the surrounding IBIs, it was expected at the big red arrow pointing downwards. Also, The QRS complex is clearly wider compared to the sinus beats.
When an ectopic beat originates above the ventricle, the electrical current will travel through the bundles of His down the ventricle, thus giving a normal narrow QRS complex. When one sees a short IBI and the P wave (representing the depolarization of the atria) has a different morphology or is not present at all, this is most likely a premature atrial contraction (PAC). Below, an example is shown.
TIP13: You can also choose to export the interbeat time interval series to a text file for use in different software packages like the CarSpan or Kubios by clicking on the menu buttons ‘Export Beats To ASCII File’. Chose the appropriate directory and file name and save the IBI time series as a text file.
VU-DAMS gives an in option in the artefact pop-up window option to mark these beats as ‘premature ventricular contraction’ or ‘premature atrial contraction’. The count of PVC’s or PAC’s per label (if present) are included in the final output.
Note that, when there is a period of a lot ectopy, you might have to delete the period entirely when computing HRV, instead of the single ectopic beats as the HRV variables cannot be calculated reliably.
Several variables are derived from the R-peaks signal: heart rate (HR), inter-beat-interval (IBI) and heart rate variability (HRV). For each variable an average per label is calculated (see section 5.2.2 Labeling your data). The values are displayed in the Results tab (section 5.7).
HR and IBI are two measures of the same construct. The HR is the number of R-peaks which occurs in one minute (beats per minute), while the IBI is the time between two consecutive beats in milliseconds. HR can be derived from IBI by the formula: HR = 60000 / IBI. HR is the oldest known index of ANS activity. Each heart beat is initiated by the sinoatrial (SA) node, the “pacemaker” of the heart. The SA node has an intrinsic rate between 60 and 100 beats per minute, varying depending on an individual’s age and gender (Kashou, Basit, & Chhabra, 2020). However, resting state HR is usually lower than this intrinsic rate due to tonic parasympathetic influences on the SA node. When PNS activity is high it acts as a “brake” on the intrinsic rate by lowering it, while this brake almost disappears when PNS activity is low (Porges, 2011). However, the rate of the heart at any given moment in time is determined by the interplay between the sympathetic and parasympathetic nervous system on the SA node. Increased SNS activity leads to an increase in HR while increased PNS activity leads to a decrease in HR (Kashou, Basit, & Chhabra, 2020; Smith, Thayer, Khalsa, & Lane, 2017). Due to the mixed effects of both SNS and PNS activity on the resulting HR, it can be used as an index of arousal but the relative contribution of these separate ANS branches cannot be detangled.
A frequently used measure of PNS activity is HRV. As its name implies it represents the variability in the time interval between successive heartbeats. Due to the fast temporal kinetics of the parasympathetic signaling (<1s) at the SA node, changes in PNS activity are detectable on a beat-to-beat scale. Sympathetic signaling on the other hand has much slower effect (>5s) and cannot change HR so quickly (Chapleau & Sabharwal, 2011). Therefore, fast changes in HR can be attributed to changes in
PNS activity. Because PNS signaling to the heart is delivered by the 11th cranial nerve, called the vagus nerve, PNS activity is often referred to as vagal tone. PNS activity can be indexed by changes in HRV in both the time domain and the frequency domain.
Time domain measures quantify the amount of variability in the IBI over time (Shaffer & Ginsberg, 2017). DAMS calculates two time-based measures of HRV: the root-mean-square of successive differences (RMSSD) and the standard deviation of the N-N interval (SDNN).
RMSSD is calculated by taking the difference between successive R-R intervals in milliseconds, squaring these differences to make negative values positive and then taking the root. The average value for all successive IBIs in a label is the final RMSSD. SDNN is calculated by taking the standard deviation of a set of IBIs in milliseconds. The term normal-to-normal (N-N) interval is just another name for the R-R interval.
Spectral analysis is performed on the IBI time series according to obtain high frequency (HF) and low frequency (LF) HRV by the following method:
The IBI time series within each label is interpolated with a cubic spline and the resulting function is resampled at 4 Hz. The resampled signal is split into overlapping periods of 256 seconds, each with 1024 data points. The overlap between two consecutive periods is 128 seconds. Periods that have intervals longer than 5 seconds without IBIs are discarded. Missing data from the final period are padded with zeros. Each period of 1024 data points is convoluted with a smoothness prior matrix (see: An advanced detrending method with application to HRV analysis, Mika
P. Tarvainen, Perttu O. Ranta-aho, and Pasi A. Karjalainen) to yield a stationary signal on which a discrete Fourier analysis is performed after additional convolution with a quadratic window. Power values for each of the 1024 data points are then averaged across all available periods in the condition (Welch method). Next the total power in the 0.0001 Hz to 0.4 Hz range is computed (TP) as well as the power in the 0.04-0.15 Hz band (LF) and the 0.15-0.40 Hz band (HF).
LF power is caused by blood pressure oscillations affecting both SNS and PNS cardiac control. HF power is caused by the effects of respiration on PNS. It is important to note that the IBI time series is first detrended and ‘corrected’ by interpolation to
deal with too short and too long IBIs (e.g. in case of an extrasystolic beat) because slow trends as well as strongly deviant beats can distort the spectrum. Because at least 3 minutes are required to obtain a reliable estimate of the LF power, these values are not supplied for labels with a duration shorter than 3 minutes.
TIP14: In case you want to change the default settings of the power spectral analysis select Edit → Settings → Frequency Analysis. Settings for preprocessing affect detrending and deviant beat removal (‘artefact’). The frequency bands reflect the typical bands now commonly used in literature, but can be changed if desired (also see appendix A).