Fig 1.
Wrong R-wave peaks detection in a pacemaker pattern.
(a) Determine pacemaker pulses as R-wave peaks; (b) Correct detection of R-wave peaks.
Fig 2.
Determination of the time delay ‘differ’ between the corresponding peaks of the ECG and BP.
(a) Illustration of the ‘differ’ between the ECG and continuous BP; (b) Cross-correlation function to determine the ‘differ’.
Fig 3.
Overall block diagram for the proposed R-wave peak detection algorithm.
Fig 4.
Flowchart of ECG detection.
Fig 5.
ECG preprocessing procedure (preprocessing-ecg).
Fig 6.
Flowchart for removing the pacemaker pulse.
Fig 7.
Flowchart of BP detection.
Fig 8.
BP preprocessing procedure (preprocessing-bp).
Fig 9.
Different kinds of qualities on simultaneously recorded ECG and BP signals.
(a) BP is not monitored; (b) (c) the BP waveform is better than the ECG waveform; (d) the ECG waveform is better.
Fig 10.
Flow diagram for fusion.
Table 1.
Results of different techniques on the updated challenge training database.
Table 2.
Results of different techniques on the MGH/MF waveform database.
Fig 11.
Comparison of detecting R-wave peaks independently and with the help of the BP signal.
(a) Marks of the ECG R-wave peaks, which were detected independently. (b) Official marks of the ECG peaks. (c) Marks of the ECG peaks using our proposed fusion algorithm. (d) Marks of the BP peaks by our proposed fusion algorithm.
Fig 12.
Successfully detecting the R-wave peaks in different patterns using our proposed algorithm.
(a) A record with noisy ECG signal; (b) A record with pacemaker; (c) A record with a sudden change of RR-intervals.
Fig 13.
Performance scores of our proposed algorithm under different heart rate variabilities and average heart rates: (a) Different HRV for Challenge database; (b) Different HRV for MGH/MF database; (c) Different average heart rate for Challenge database; (d) Different average heart rate for MGH/MF database.
Table 3.
Performance scores of challenge database grouped by RMSSD.
Table 4.
Performance scores of MGH/MF waveform database grouped by RMSSD.
Table 5.
Performance scores of challenge database grouped by average heart rate (HR).
Table 6.
Performance scores of MGH/MF waveform database grouped by average heart rate (HR).
Fig 14.
Cases of missed or redundant detections.
(a) A missed mark in ECG; (b) A redundant mark in ECG; (c) Poor detecting results when the current ECG and BP segments are both of poor quality;(d)Inconsistent detecting results with the wrong official reference annotations, in which the * is the correct annotation, and the ⃝ is the annotation using the proposed algorithm.