This work is licensed under the Creative Commons Attribution 4.0 International License.
Figure 1
Placement of Shimmer3 measurement units (white boxes), ECG electrodes (blue circles) and the PPG sensor (green box).
Figure 2
A seven beat PPG sequence showing clipping of the systolic peaks (blue curve) and the spline-based interpolation used (orange curve). The two fiducial points assessed is also marked in the figure.
Figure 3
Example showing 13 beats of PPG recording together with the detected foot (red crosses) and max slope (black crosses) fiducial points. The foot points were located based on identifying negative-to-positive zero-crossing points of the first derivative of the signal, and the subsequent peak of the derivative was used to locate the max slope fiducial point.
Figure 4
Average heart rate for all participants during the different periods.
Figure 5
Example segments from recordings with good PPG signal quality (a), during the resting phase, and with poor PPG signal quality (b) during the exercise phase.
Figure 6
A transient disturbance in the PPG signal (upper plot) causing a large deviation in the estimated inter-beat interval over two points (blue line in lower plot). The red line shows the IBI time-series after outlier correction. The red and green crosses in the upper plot mark the foot and maximum slope fiducial points, respectively.
Figure 7
The spread of SDNN for the 15 participants measured with ECG and PPG with and without outlier correction, broken down by the different phases and fiducial points (foot and maxdiff). Outliers are shown in red.
Figure 8
The spread of RMSSD for the 15 participants measured with ECG and PPG with and without outlier correction, broken down by the different phases and fiducial points (foot and maxdiff). Outliers are shown in red.
Figure 9
The difference between SDNN values as calculated from PPG and ECG for the 15 participants in the different phases. Grouped by fiducial point and outlier correction. Outliers and median values are shown in red.
Figure 10
The difference between RMSSD values as calculated from PPG and ECG for the 15 participants in the different phases. Grouped by fiducial point and outlier correction. Outliers and median values are shown in red.
Summary of agreement between HRV derived from ECG and PPG for each type of activity, depending on PPG fiducial point selection with and without correction of outliers. The agreement is represented as median absolute deviation (MAD), median absolute relative deviation (MARD) and the root mean square error (RMSE).
Fiducial point
Outlier correction
HRV
Agreement
Relaxation
Exercise
Mental stress
Foot
Yes
SDNN
MAD
9,83
137,86
10,14
Max slope
Yes
SDNN
MAD
2,25
78,82
1,19
Foot
No
SDNN
MAD
28,12
290,49
26,95
Max slope
No
SDNN
MAD
21,98
272,73
17,6
Foot
Yes
RMSSD
MAD
27,56
189,3
50,17
Max slope
Yes
RMSSD
MAD
3,46
114,26
2,27
Foot
No
RMSSD
MAD
64,21
415,58
72,18
Max slope
No
RMSSD
MAD
41,31
369,77
27,77
Foot
Yes
SDNN
MARD
18,00 %
341 %
23,20 %
Max slope
Yes
SDNN
MARD
3,60 %
239 %
1,80 %
Foot
No
SDNN
MARD
51,50 %
933 %
53,90 %
Max slope
No
SDNN
MARD
22,00 %
877 %
28,30 %
Foot
Yes
RMSSD
MARD
85,00 %
1576 %
156 %
Max slope
Yes
RMSSD
MARD
8,20 %
647 %
6,40 %
Foot
No
RMSSD
MARD
152 %
1727 %
216 %
Max slope
No
RMSSD
MARD
60,00 %
1572 %
79,30 %
Foot
Yes
SDNN
RMSE
21,41
202,95
27,33
Max slope
Yes
SDNN
RMSE
14,83
184,50
8,59
Foot
No
SDNN
RMSE
85,16
403,71
60,56
Max slope
No
SDNN
RMSE
79,17
395,91
49,43
Foot
Yes
RMSSD
RMSE
43,17
253,67
66,01
Max slope
Yes
RMSSD
RMSE
14,98
212,15
22,28
Foot
No
RMSSD
RMSE
143,70
552,46
126,00
Max slope
No
RMSSD
RMSE
131,14
533,94
99,57
Summary of agreement between RR derived from ECG and IBI derived from PPG for each type of activity. The agreement is presented as median absolute deviation (MAD), median absolute relative deviation (MARD) and the root mean square error (RMSE).