Accurate neonatal heart rate monitoring using a new wireless, cap mounted device

A device for newborn heart rate (HR) monitoring at birth that is compatible with delayed cord clamping and minimises hypothermia risk could have advantages over current approaches. We evaluated a wireless, cap mounted device (fhPPG) for monitoring neonatal HR.

interventions, and international guidelines suggest the electrocardiogram (ECG) may be used in the delivery room to monitor it. 4 However, electrode application can be difficult, ECG monitors are not always available and electrical cardiac activity may not always equate to effective cardiac output. 5,6 Pulse oximetry (PO) is widely used on neonatal intensive care units (NICUs), but uptake in the delivery room is not universal 7 potentially because of the delay of a few minutes to obtain a reliable HR after birth due to poor peripheral perfusion and motion. 8 This could explain why PO underestimates HR when compared to ECG during the first minutes of life. 9,10 Delayed cord clamping (DCC) for term infants is widely practised and could become more common for preterm infants, but there is a need to better monitor these babies during this transition to maximise the benefits of DCC whilst avoiding any adverse effects of delayed resuscitation. 11,12 Maximising placental transfusion requires the attending team to be confident of the baby's condition including an acceptable HR and avoidance of hypothermia. A non-intrusive, wireless, easy to apply HR monitoring system, which avoids the risk of hypothermia, could be advantageous in this setting.
We have previously described a prototype wired, forehead-mounted sensor studied in NICU patients that utilises reflectance-mode green light photoplethysmography (PPG) to measure HR. 13 This has the advantages that this wavelength of light (525 nm) is optimised for reflection-mode detection of blood flow and captures this at the forehead where brain perfusion shares arterial pathways via the carotid artery and so is less susceptible to poor peripheral perfusion. 14,15 This original device was held in place on the forehead by a spun bond laminate headband and successfully demonstrated that HR measurement by this method was feasible but had a Bland-Altman limit of agreement (LOA) of up to ± 12bpm. In addition, it was not compatible with current neonatal care due to its inability to accommodate endotracheal tube and CPAP attachments. Furthermore, the relatively large sensor (22 mm diameter), large electronic circuit boards and the presence of cumbersome wiring were less suited for bedside stabilisation of preterm infants during DCC. This device has now been substantially modified (Video S1, Figure S1 and Figure S2) in terms of its practicality, and several major modifications have taken place including the following: (i) A T-shape cap compatible with respiratory equipment fittings.

| Study design
There were two phases to this study, the NICU and ECS. On the NICU, an appropriately sized fhPPG cap was fitted to the infant's head and connected wirelessly to a bespoke synchronised data-

Key notes
• Wireless, continuous heart rate (HR) monitoring of newborns using a cap mounted device (fhPPG) could be beneficial during delayed cord clamping, especially in preterm infants.
• Using electrocardiogram HR as a gold standard, a novel fhPPG performed similarly to pulse oximetry in the neonatal intensive care and immediately after caesarean section birth.
• The fhPPG is an accurate, wireless alternative to current methods of HR monitoring in newborn infants.
to 30 minutes of raw physiological data were collected with the infant in their cot or incubator.
During the ECS phase, the same monitoring system described above was deployed. Following birth, the baby was shown to parents before being placed on the resuscitaire. All monitoring equipment was attached in the same sequence for each patient by the research team. The fhPPG cap was fitted, ECG leads attached and a PO sensor attached to the wrist of the right hand [16][17][18] and raw data collected for up to 20 minutes.
The B450 monitor generates HR data which provides ECG and PO values every 5 seconds. All fhPPG HR data extracted were averaged over the same 5 seconds window for direct comparison.
Slow and rapidly changing HRs can be unpredictable in the NICU and ECS settings. To ensure the fhPPG HR algorithm can effectively extract a wide range of rapidly changing HRs, as experienced during newborn care, we subjected the algorithm to simulated HR data covering 25-250 bpm using the Japanese Industrial Standard (JIS T 1303:2005, revised 2018). 19 Although the JIS is designed for testing foetal heart rate monitors, it appropriately models newborns as the HR baseline and variation are similar to those of the foetus. 20 A simulated PPG signal was input across the JIS range of HRs, the algorithm's performance tracked and linear regression applied to evaluate the overall performance.

| RE SULTS
A convenience sample of 60 infants was recruited (NICU = 40, ECS = 20) when researchers were available. Six infants were excluded from the NICU phase and two from the ECS phase due to ECG data not being stored or corrupted leading to data loss and misalignment. The patient demographics can be seen in Table 1.

| Success rate
During the NICU phase, the median ECG success rate was 100% (IQR,  97.2-100, n = 1684). Device comparison with the gold standard was also determined using both Bland-Altman plots (  Figure 1).

Linear regression of all pooled paired HRs demonstrated PO
and fhPPG had similar goodness-of-fit with both having an R 2 = .89 ( Figure 2). In five babies, the ECG measured a HR <100 bpm. For these, there were 23 paired values with the PO of which two reported a HR >100 bpm and for the fhPPG there were 13 paired values with two reporting a HR >100 bpm, and on both occasions the fhPPG output was identified as a poor signal triggering a warning to the user.

| Algorithm simulation
Simulated PPG data were applied as described across a range of HRs and with varying rates of change as defined by JIS T (Figure 3). The fhPPG algorithm demonstrated excellent correlation with the simulated HR (Figure 4).  It is essential that any newborn HR monitoring device is able to detect low rates, particularly those below 100 bpm. The majority of babies in this study were stable with limited HR variability so limiting the number of HRs <100 bpm. Use of the JIS methodology allowed us to test the software algorithm to rapidly changing simulated HRs across the typical newborn range. The fhPPG algorithm was able to track the HRs quickly and accurately with an R 2 = .994. This strong correlation supports the accuracy of the device's software based on simulation data, but the whole system requires real-world evaluation with infants undergoing resuscitation or stabilisation.

| D ISCUSS I ON
The secondary aim of this study was to explore the acquisition Whilst fast detection of electrical activity of the heart is useful, when cardiac output is poor or non-existent 5,6 there could be advantages to having a parallel optical sensor that is able to detect blood flow and hence confirm the electrical activity is associated with cardiac output.
Any potential benefits of a centrally placed, green light optical sensor, that is fhPPG which detects blood flow from the forehead sharing a common central arterial blood supply to the brain, compared to a peripherally sited transmission mode device, require further evaluation in larger number of patients including preterm infants.

| CON CLUS ION
With the increasing use of DCC, the development of a wireless, continuous HR monitor that aligns with the normal care pathway, including placement of a cap with attachment for respiratory equipment, could better support transitioning of newborn babies and audit interventions. 25 Confidence in such a device could allow monitoring of the newborn and aid decisions around when to clamp the cord and instigate any stabilisation or additional resuscitation steps.
fhPPG operates in reflectance mode allowing detection of a PPG signal on many parts of the body. The forehead allows inclusion in a cap, normally applied to keep the baby warm, and could potentially improve reliability during low perfusion states with a shared carotid artery blood supply to the brain. 13 The wireless design avoids multiple trailing wires that could become accidentally detached during bedside care with DCC. This device provides healthcare professionals with an accurate alternative to current neonatal HR monitoring.
Further development of the HR algorithm, addition of oxygen saturation monitoring and evaluation during low perfusion states are underway.

ACK N OWLED G EM ENTS
Dr Andrew Prayle helped to implement the modified Bland-Altman for multiple measurements. Thank you to all the families who kindly agreed to participate in this study.