Skin-interfaced biosensors for advanced wireless physiological monitoring in neonatal and pediatric intensive-care units

Abstract

Standard clinical care in neonatal and pediatric intensive-care units (NICUs and PICUs, respectively) involves continuous monitoring of vital signs with hard-wired devices that adhere to the skin and, in certain instances, can involve catheter-based pressure sensors inserted into the arteries. These systems entail risks of causing iatrogenic skin injuries, complicating clinical care and impeding skin-to-skin contact between parent and child. Here we present a wireless, non-invasive technology that not only offers measurement equivalency to existing clinical standards for heart rate, respiration rate, temperature and blood oxygenation, but also provides a range of important additional features, as supported by data from pilot clinical studies in both the NICU and PICU. These new modalities include tracking movements and body orientation, quantifying the physiological benefits of skin-to-skin care, capturing acoustic signatures of cardiac activity, recording vocal biomarkers associated with tonality and temporal characteristics of crying and monitoring a reliable surrogate for systolic blood pressure. These platforms have the potential to substantially enhance the quality of neonatal and pediatric critical care.

Fig. 1.. Designs, mechanical characterization results and photographs of a soft, wireless chest unit for physiological monitoring of neonatal and pediatric patients.

a, Schematic diagram and expanded view illustration of a device with a modular primary battery. The main body consists of buckled serpentine interconnects between islands of electrical components contained within a soft, elastomeric enclosure. The battery interfaces to the system via reversible magnetic coupling. Thin silicone pads establish electrical connections between measurement electrodes and a hydrogel interface to the skin, to yield a completely sealed, waterproof device. LMS stands for low modulus silicone. b, Illustration of a detachable wireless power harvesting system. c, Illustration of a powering option that involves an integrated, wirelessly rechargeable battery. This option uses a different top layer encapsulation, without the magnets. d, Photograph of the chest unit with modular battery on a realistic model of a neonate. e, Computed stresses (right: normal; left: shear) at the interface between the skin and a device during uniaxial stretching to a strain of 20%, with a thickness of 200 μm. f-h, Photograph of a representative device during stretching (f), twisting (g), and bending (g). i-j, Photographs of a device with integrated battery (i) and with a wireless energy harvester (j), both mounted on a model.

Fig. 2.. Designs and photographs of a wireless limb unit for physiological monitoring of neonatal and pediatric patients and block diagram of the system operation.

a, Schematic diagram and expanded view illustration of the limb unit, designed to measure PPG, SpO₂ and peripheral skin temperature. b, Photograph of a device while bent and twisted. c-e, Placement of a device on (c) a model of a neonate at the ankle-to-base of the foot, (d) a model of a pediatric patient (2-year-old CA) at the wrist-to-hand, and (e) at the foot-to-toe. f, Photograph of the chest and limb unit on a model of a neonate in a NICU isolette, with a tablet computer displaying representative data through a graphical user interface. g, Block diagram showing the operational scheme of two time-synchronized devices, with an analog-front-end for ECG processing, 3-axis accelerometer, thermometer IC, and BLE SoC for the chest unit and a pulse oximeter IC, thermometer IC, and BLE SoC for the limb unit. Three different options for power supply appear at the bottom of g.

Fig. 3.. Photographs of wireless wearable devices on neonatal and pediatric patients in the NICU and PICU, respectively, and of parental hands-on care with healthy neonates.

a-c, Photographs of the chest unit on two different ~2-year-old children (a) and (b) and on a neonatal preterm infant in (c) (27 w GA, 6 w CA) with apnea of prematurity and respiratory failure. Here, the device rests on the back. d, Photograph of the limb unit mounted on the ankle-to-foot interface of an infant (27 w GA, 6 w CA) and e, another infant (25 w GA, 44 w CA). f, Photograph of a similar unit on the foot-to-toe interface of the pediatric patient in (b). g, Photograph of a device on the wrist-to-hand interface of a pediatric patient (35 w GA, 6w CA). h-j, Photograph of a pair of devices on a 40 w GA healthy neonate during (h) KC, (i) feeding, and (j) hands-on care (diaper change).

Fig. 4.. Representative data collected in the NICU and PICU.

a, Representative ECG, PPG, SCG and respiration waveforms collected from a neonate (29 w GA). b, Comparison of HR, SpO₂, RR, temperature, and temperature gradient between the chest and the foot to standard clinical measurements. c-f, Corresponding Bland-Altman plots for (c) HR, (d) SpO₂, (e) RR, and (f) temperature.

Fig. 5.. Time-synchronized operation of chest and limb units for measurements of systolic blood pressure, with comparison to arterial line data collected in the PICU on pediatric patients.

a, Block diagram of the scheme for time-synchronization. b, Definition of pulse arrival time (PAT) and pulse transit time (PTT), as derived from ECG, SCG, and PPG waveforms. PEP and AO stands for pre-ejection period and aorta opening, respectively. c, Representative PAT and PTT data from a pediatric patient in the PICU during a study over 5 h. d, Results of SPB determined with PAT and PTT and with an arterial line (A-line) for an infant (34 w GA, 40 w CA) and e, another infant (40 w GA, 50 w CA). f,g, Bland-Altman plot for PAT-derived SBP (n = 5) and Bland-Altman plot for PTT-derived SBP (n = 5), respectively. Data points in the red circle indicates the comparison result at the incidents of motion artifact in 5e.

Fig. 6.. Advanced monitoring modalities based on measurements of orientation, activity and vibratory motions.

a, Definition of device axes and rotation angle between the device and reference frames. b, Orientation data extracted from low bandpass filtering of accelerometry data collected from the chest unit and derived rotation angles for various scenarios: resting in supine and right lateral positions, non-KC and KC holding. c, Filtered accelerometry data and d, rotational angles for neonates in various body positions in the NICU (n = 3). e, Representative HR, SpO₂, chest and peripheral limb temperature, orientation and activity data defined as the root-mean-square of the 3-axes acceleration values between 1 and 10 Hz before, during, and after KC with a premature neonate (31 w GA). f, Raw accelerometry signal (top) and spectrogram of time-frequency signal (bottom) during crying and non-crying events from a neonate (37 w GA, large-for-gestational-age) with feeding difficulties. g, Comparison of cry duration determined with the device and by manual recording from neonates (n = 3), with a total of 11 cry events. The resolution of cry duration from our device was 0.2s, whereas manual count ranged from one second to one minute, dependent on the recorder. Error bars were defined as half of the resolution of the measurement.