A Wearable Sensor for Monitoring KangarooMother Care treatment for premature neonates

Ashish JoglekarRBCCPS, IISc

Bangalore, Indiaashishj@iisc.ac.in

Alok RawatRBCCPS, IISc

Bangalore, Indiaalokrawat@iisc.ac.in

Vasanth RajaramanRBCCPS, IISc

Bangalore, Indiavasanthr@iisc.ac.in

Bharadwaj AmruturRBCCPS, IISc

Bangalore, Indiaamrutur@iisc.ac.in

Prem MonySt. John’s Research Institute

Bangalore, Indiaprem mony@sjri.res.in

Prashanth ThankachanSt. John’s Research Institute

Bangalore, Indiapaxmail@gmail.com

Tony RajSt. John’s Research Institute

Bangalore, Indiatonyraj@sjri.res.in

Suman RaoSt. John’s Medical Hospital

Bangalore, Indiasuman.rpn@stjohns.in

Abstract—We describe a new wearable device that has beendeveloped to monitor the effectiveness of Kangaroo MotherCare (KMC) treatment administered to premature neonates. Weidentify the sensors needed to monitor KMC treatment. Ourdevice measures the neonate’s and caregiver’s skin temperatureand the neonate’s relative position during KMC. The deviceincorporates skin touch sensors to qualify the temperaturereadings, thus preventing false alarms. Design constraints relatedto form factor, placement and battery life and few measurementresults from pre-clinical trials are provided.

Index Terms—wearable; premature neonate; kangaroo mothercare; skin temperature sensor; touch sensor; tilt sensor

I. INTRODUCTION

Neonatal Mortality Rate (NMR) quantifies the number ofnew-born deaths (that is in the first month of life) per thousandlive births and is responsible for more than 46% of all childdeaths before the age of five (2.6 million in 2016) [1]. Morethan 77% of these deaths occur in developing nations. Reports[1], [2] have shown that India has about 10 times higher NMRcompared to the western world. NMR in India was 25 as of2016, and has reduced by 50% since 1990. However, consid-ering India’s burgeoning population, 6, 40, 000 newborns diedin India in 2016, which is about 8 times more than in China[1]. The authors in [3], [4] have found hypothermia to be amajor cause of neonatal morbidity and mortality, especiallyin rural and other resource constrained settings. Hypothermiafor neonates is defined as the condition when their bodytemperature drops below 36.5C [5]. Hypothermia leads togreater risk of infection and may even lead to fatality in theabsence of pre-emptive care.

Premature neonates are usually treated in an incubatorin an NICU (Neonatal Intensive Care Units). The incubatoressentially has heating lamps and other electronics that helpsto keep the baby at the right temperature. This treatment canlast anywhere from a few days to a few weeks, based ontheir degree of prematurity. Unfortunately, in many resourceconstrained settings, especially in the rural areas, there are notenough NICUs or even incubators available thus preventing thedelivery of this essential treatment to the premature infant.

Fortunately, a simple and effective treatment was discov-ered in early 1980s by Dr. Sanabria [6], who, inspired byKangaroos, found that holding the premature infant in skinto skin contact with the caregiver, resulted in very dramaticimprovements in the prognosis for the premature infant. Thistechnique is now called Kangaroo Mother Care (KMC) and isbeing rapidly adopted by several hospitals. Given that almostno other equipment is required, this seems to be very suitablefor use in resource poor settings. A recent study by Charpaket. al. [7] has further highlighted the long-lasting social andbehavioural protective effects in the original set of patientsfrom Columbia, even 20 years after the initial intervention.

Ideally, during this treatment, the doctors would like tomonitor the temperature of the baby, whether the KMC isbeing applied and if so for what duration in a 24-hour periodand whether the baby is being held in the correct position.Such data would help the doctor to track the progress ofthe health of the baby and correlate with the treatment givenand take necessary actions as required. However, there is noexisting technology currently to quantify the application ofKMC treatment, other than manual recording by the care-givers, which is cumbersome and prone to errors.

In response to this need, we have developed a new wearabledevice to continuously monitor and record the KMC parame-ters. The parameters measured are baby and caregiver’s skintemperature (qualified by skin touch sensors for both) and thebaby’s relative position. We found that the wearable’s allow-able form factor (particularly device thickness that ensures usercomfort) placed a severe constraint on the battery size limitingthe achievable battery life . Choice of low power sensors andduty cycling of the wireless radio helped us meet the batteryendurance requirement (KMC treatment period of four weeks).The constrained form factor meant that the sensors had tobe packaged in very limited volume without compromisingperformance (due to issues like crosstalk). As the device getssandwiched between the baby and caregiver when in KMC, theantenna placement also had to be optimised to meet wirelessrange and regulatory requirements (skin absorption rate).

Fig. 1. Remote Kangaroo Mother Care monitoring system

By using the IoT (Internet of Things) approach, informationfrom the wearable is available almost instantaneously at theremote hospital (provided network connectivity is present).Simple automated processing of the data is done to raise analarm in case of adverse changes in the baby’s temperature,so that appropriate intervention can be taken immediately.

II. END TO END SYSTEM FOR KMC MONITORING

The basic concept of the end-to-end system is illustratedin Fig. 1. The key component of the system, is the wearabledevice shown in the picture on the abdomen of the baby. Thedevice is meant to be affixed around the navel of the baby,from where it picks up the baby’s skin temperature. Since thisposition is also close to the liver, it gives a good approximationof the baby’s core temperature. There is another temperaturesensor on the opposite side of the device, which can pick upthe temperature of the caregiver. When the caregiver holdsthe baby to provide skin to skin contact, the device getssandwiched between the baby and the caregiver and touchsensors are used to signal the onset of the KMC treatment.During KMC, skin touch sensors on opposing surfaces of thedevice validate that the temperature being measured is that ofthe human body. An accelerometer also measures the angle atwhich the baby is held when in KMC. The device measures thetime for which the treatment is given. Thus, the device picksup all the parameters necessary for monitoring of the KMCtreatment. The device automatically and wirelessly connects(over BLE) with a smart phone. The smart phone uses themobile network to transfer this data to the hospital’s server.The device can store 30 hours of data sampled every 6 minutes.Technical details of the device’s design and construction arediscussed in the next section.

The device dimensions are 3cm × 1.5cm with a thicknessof 7mm. The dimensions were optimised to prevent anydiscomfort to baby or care giver and were arrived at basedon feedback obtained from users of the device. The device ishoused in a bio-compatible plastic shell and has two small,medical grade, stainless steel windows on each surface. ICUgrade NTC type thermistors [9] attached to these windows

Fig. 2. Exploded view of KMC device.

sense the temperature on both sides of the device (baby andcare giver side).

Alarms are provided in the system to detect various condi-tions like mild hypothermia 34.1C and 36.5C, severe hy-pothermia < 34.1C or failure to administer KMC treatmentin a six hour period.

III. KMC DEVICE DESIGN

A. Hardware Design

Fig. 2 shows the hardware implementation of our KMCdevice. Also indicated are the types and locations of sensorsin this device.

1) Temperature Sensors: Two ICU grade NTC type ther-mistors are used for sensing temperature on baby and caregiver side. The sensors come calibrated at 37C ±0.01% andare grouped according to their resistance range to provideextreme interchangeability [9]. The thermistor’s RT character-istics are shown in Fig. 3. R is the resistance of thermistor attemperature T , R0 = 29, 700Ω is the calibrated resistance ofthermistor at temperature T0 = 37C ±0.01%, β = 3950K isthe material specific constant. The linear fit reveals the sensor’ssensitivity to be ' −1768.1Ω/C. The thermistor resistanceis measured in comparison to a reference grade resistor. Theentire analog front end, consisting of the resistor network andADC are powered by the same switchable power supply tominimize the impact of supply noise [8].

2) Skin Touch Sensors: Skin touch detection is based onmeasurement of change in capacitance when the device isbrought in proximity of the human skin. The KMC devicebeing battery operated has no reference to physical ground.Thus, a human body in proximity of the device will be ata different potential considering our battery-operated device’sfloating ground reference. As a result, the human body cannotbe used as a reliable second plate of a capacitor as is the case inthe traditional approach to capacitive touch sensing. Instead aground plane is kept in close proximity of the capacitive touchelectrode. Proximity to a human body changes the dielectric

Fig. 3. ICU grade NTC thermistor’s Resistance(Ω) vs Temperature(C) plot.

constant across these two plates thus changing the capacitance[9].

The onboard sensor controller measures capacitance bycalculating the time it takes to charge this capacitor on passinga constant current of 1µA. We have placed two such touchsensors on the baby and care giver side of the device. Whenthe device was made thinner to meet the 7mm thickness spec-ification, the two touch sensors began to experience crosstalk(touch was detected on both sides though only one side wasbrought in proximity of human skin). Placement of a shieldedground plane between the two electrodes helped correct thiscrosstalk issue. No explicit ground plane was introduced,instead a piezo buzzer was sandwiched between the two touchsensing surfaces (Fig. 2). The conductive surface of the buzzerprovided enough shielding to prevent crosstalk.

3) Position Sensor: The device makes 3 axis accelerometermeasurements using the onboard Invensense 9250 MEMSbased Motion Processing Unit (MPU). The accelerometer’sfull scale range is set to ±2g. Data is captured at 16 bitresolution, with a sensitivity of 16, 384LSB/g. The sensitivitydrifts with temperature at a rate of ±0.026%/C. The X, Ycomponent level initial zero-g calibration tolerance is ±60mg.The corresponding tri-axis tilt angles are calculated using basictrigonometric equations to determine tilt angle of the baby.

4) Antenna Position: A 2.4GHz BLE ceramic mini an-tenna is placed on care giver side of the PCB. The TX poweris set at 0dBm which corresponds to a Class 3 device (peaktransmit power of 1mW ). The effective exposure (SAR) is0.00117–0.00319W/kg is well below 2W/kg recommendedby the ICNIRP [10]. The metallic battery enclosure and coincell between antenna and baby provide further reduction inexposure on baby side.

5) Power Supply: The device is powered with a singleCR2025 coin cell. The cell capacity is ∼ 170mAh. TheBLE advertisement and sensor controller sampling rate is dutycycled to achieve a battery life of more than 30 days. An onchip internal DC-DC converter provides a regulated DC bus.

B. Product Design

Fig. 2 shows an exploded view of the KMC Sensor device.The temperature sensors are attached to the inner side oftwo biocompatible stainless steel (SS) plates using thermallyconductive adhesive. Two additional copper pads are bedded

in the inner surface of the shell to form the two platesof a capacitor for the touch sensor. Since these plates arecovered with the insulating plastic enclosure and do notmake direct physical contact with the skin, we have chosenhighly conductive copper as the material for these plates. Thissolution is low cost (avoids need for bio compatible touchsensor metal plates) and easy to assemble considering the easeof soldering a wired contact. All surfaces are bio-compatibleand hypoallergenic. The PCB board houses the sensor analogfront end, microcontroller, ADC, storage, position sensor andBLE communication module. A standard coin cell is used asthe power source. The device may be affixed to the subjecteither via a belt, a tape or a bandage. The device also hasa multiplicity of LEDs of different colors, that are used toindicate various events or conditions to the user. The devicealso has a piezo buzzer to emit various sounds upon differentevents or conditions.

IV. MEASUREMENT RESULTS

This device was used for a pre-clinical study for about11 months (Feb – Dec 2016) and has been used to monitorKMC on 73 babies during this period. Initial analysis ofthe field data is reported in [11]. Fig. 4 shows data fromthe device in comparison to standard manual method ofmeasuring temperature from the axilla (armpit), taken from the73 babies during the home use phase. The data is plotted usingthe Bland-Altman style plot and shows the tight correlationbetween the device data and that measurement using thestandard practice. Another interesting observation from thisstudy has been that the self-reporting of KMC duration hasbeen consistently higher than what is measured.

V. SUMMARY

The new wearable sensor for measuring KMC dosage forpremature neonates has shown promising results in field test-ing. The device and system design incorporate many featuresfor robust measurement and retrieval of data. A key featureis to predicate the readings of the internal sensors with theskin-touch sensor in order to minimize false alarms.

ACKNOWLEDGMENT

The authors are grateful to the nursing team at St. John’sHospital and St. John’s Research Institute. This work wasfunded by a grant from the Bill and Melinda Gates Foundation.

Fig. 4. Bland-Altman plot of device vs manual-axila readings. SD of ±0.3Cindicates good precision and the mean ' 0.1C indicates good accuracy [11]

REFERENCES

[1] Hug, L., D. Sharrow, and D. You, “Levels and trends in child mortality:report 2017. Estimates developed by the UN Inter-agency Group forChild Mortality Estimation.” (2017).

[2] Million Death Study Collaborators, “Causes of neonatal and childmortality in India: a nationally representative mortality survey.” TheLancet 376, no. 9755 (2010): 1853-1860.

[3] Kumar, Vishwajeet, Saroj Mohanty, Aarti Kumar, Rajendra P. Misra,Mathuram Santosham, Shally Awasthi, Abdullah H. Baqui et al. “Effectof community-based behaviour change management on neonatal mor-tality in Shivgarh, Uttar Pradesh, India: a cluster-randomised controlledtrial.” The Lancet 372, no. 9644 (2008): 1151-1162.

[4] Kumar, V., J. C. Shearer, A. Kumar, and G. L. Darmstadt, “Neonatalhypothermia in low resource settings: a review.” Journal of Perinatology29, no. 6 (2009): 401.

[5] World Health Organization, “Thermal protection of the newborn: apractical guide.” (1997).

[6] Rey, E. S., and H. G., “Manejo racional del nino prematuro.” Proceedingsof the Conference Curso de Medicina Fetal, Bogota, Universidad Na-cional (1983): 23-25. (Spanish). (Manuscript available in English fromUNICEF, 3 UN Plaza, New York, NY: 10017).

[7] Charpak, Nathalie, Rejean Tessier, Juan G. Ruiz, Jose Tiberio Hernan-dez, Felipe Uriza, Julieta Villegas, Line Nadeau et al. “Twenty-yearfollow-up of kangaroo mother care versus traditional care.” Pediatrics(2016): e20162063.

[8] Rao, Hiteshwar, Dhruv Saxena, Saurabh Kumar, G. V. Sagar, BharadwajAmrutur, Prem Mony, Prashanth Thankachan, Kiruba Shankar, SumanRao, and Swarnarekha Bhat. “Design of a Wearable Remote Neona-tal Health Monitoring Device.” In International Joint Conference onBiomedical Engineering Systems and Technologies, pp. 34-51. Springer,Cham, 2014.

[9] B. Amrutur et. al., Device for Neonatal Monitoring, Indian PatentApplication Number: 462/CHE/2015.

[10] ICNIRP Guidelines, “Guidelines for limiting exposure to time varyingelectric, magnetic and electromagnetic fields (up to 300GHz)”, Healthphysics 74, no. 4 (1998): 494-521.

[11] Rao, Suman, Prashanth Thankachan, Bharadwaj Amrutur, MaryannWashington, and Prem K. Mony. “Continuous, real-time monitoring ofneonatal position and temperature during Kangaroo Mother Care usinga wearable sensor: a techno-feasibility pilot study.” Pilot and feasibilitystudies 4, no. 1 (2018): 99.