Cyber-Physical Platform for Preeclampsia Detection
CCyber-Physical Platform for PreeclampsiaDetection
Iuliana Marin − − − , Maria IulianaBocicor − − − , and Arthur-Jozsef Molnar − − − SC Info World SRL, Bucharest, Romania { iuliana.marin,iuliana.bocicor,arthur.molnar } Abstract.
Hypertension-related conditions are the most prevalent com-plications of pregnancy worldwide. They manifest in up to 8% of casesand if left untreated, can lead to serious detrimental effects. Early detec-tion of their sudden onset can help physicians alleviate the condition andimprove outcomes for both would-be mother and baby. Today’s preva-lence of smartphones and cost-effective wearable technology provide newopportunities for individualized medicine. Existing devices promote hearthealth, they monitor and encourage physical activity and measure sleepquality. This builds interest and encourages users to require more ad-vanced features. We believe these aspects form suitable conditions tocreate and market specialized wearable devices. The present paper de-tails a cyber-physical system built around an intelligent bracelet for mon-itoring hypertension-related conditions tailored to pregnant women. Thebracelet uses a microfluidic layer that is compressed by the blood press-ing against the arterial wall. Integrated sensors register the waveformand send it to the user’s smartphone, where the systolic and diastolicvalues are determined. The system is currently developed under Euro-pean Union research funding, and includes a software server where datais stored and further processing is carried out through machine learning.
Keywords:
Cyber-physical system, Smart bracelet, Biosensors, Bloodpressure, Hypertension.
Hypertensive disorders affect up to 8% of all pregnancies [13], with increasedprevalence in women who already carried a preeclamptic pregnancy. In theUnited States, hypertension related conditions affect between 7 to 15% of preg-nancies [15]. They are classified into chronic hypertension, gestational hyperten-sion and preeclampsia [17,26], and represent the most common medical problemencountered during pregnancy. Chronic hypertension is present before 20 weeksof pregnancy, or in cases where the woman is already on anti-hypertensive medi-cation. Gestational hypertension is diagnosed after 20 weeks of pregnancy. Whencombined with proteinuria (the presence of protein in urine), or organ dysfunc-tion such as renal or liver involvement, it is diagnosed as preeclampsia. If left a r X i v : . [ c s . C Y ] S e p Marin et al. untreated, preeclampsia can lead to adverse effects for both the mother andbaby, including restricted fetal growth, organ damage, and seizures (eclampsia),which necessitate delivering the baby to prevent further consequences [17].Early detection allows medical intervention with the aim of maintaining bloodpressure under a safe threshold, managing the time of delivery for minimal riskto both mother and baby as well as prophylactic use of magnesium sulphateduring labor [24,8].In the last decade, medicine is going through an important transformationfueled by the ubiquity of wearables, together with increased awareness and in-terest from the general public [14]. We believe this can be harnessed to improvethe outcomes for conditions such as preeclampsia, which are well known, but un-der reported in many parts of the globe. In one region, 77% of pregnant womenaffected were unaware of the condition or its possible consequences [25].In this paper we detail a cyber-physical system that targets early detectionof preeclampsia and other hypertension conditions related with pregnancy [10].The system is being developed under funding from the European Union. It inte-grates a custom developed wearable bracelet for real-time measurement of bloodpressure with a smartphone application and a software server.The system connects the would-be mother with her clinician and provides anintegrated platform to alert clinicians at the onset of persistent hypertension [20].As preeclampsia can have sudden onset and is difficult to manage, registeringits early warning signs allows more flexibility in its diagnosis and management[24].One of the principal innovations regards the sensor-driven bracelet used tomonitor blood pressure. Current products are based on infrared or oximeter sen-sors that directly measure pulse and blood oxygen levels, which they then cor-relate with blood pressure [16,3,4]. They enable taking measurements at shorterintervals and without so much preparation [3]. However, these devices can re-quire calibration and they use approximation to estimate blood pressure. Theproposed solution increases accuracy by measuring blood pressure directly andeliminates calibration requirements.The wrist-worn bracelet is built around a diminutive sensor comprised of amicrofluidic layer located between two sensing membranes. External pressureof the blood pushing on the arterial wall is picked up by the microfluidic layerand results in a change of impedance which leads to the complete signal of thewaveform being recorded. Readings are transmitted to the wearer’s smartphoneusing an integrated wireless module with a BLE chip. The energy efficiency ofBLE and the sensor allow the wrist bracelet to measure blood pressure severaltimes during an hour and transmit data without requiring frequent charging.Reading history and trends can be consulted using the smartphone application.They are also forwarded to the software server, which clinicians and end-userscan access over the web. When hypertension is detected, the associated clinicianreceives an application alert; this helps with continuously monitoring patientsand lower the time to intervention. yber-Physical Platform for Preeclampsia Detection 3
Hypertension is a major cause of premature death worldwide [27] and an im-portant cause of complications during pregnancy [2]. The idea of continuousmeasurement of blood pressure outside of medical facilities has only recentlybeen adopted and during the past years various applications and devices havebeen developed for this purpose. One such device is H2-BP, a wearable bloodpressure monitor in the form of a light wrist band [9]. It measures blood pres-sure with an accuracy of 5mmHg every 30 - 50 seconds, as well as heart ratewith an accuracy of ± Marin et al.
This application includes a maintained database comprising gathered records.The caregiver is alerted if the expectant mother is in a critical circumstance.A device utilized for dealing with the health and well-being during pregnancyis the Ava bracelet [1]. It determines cycle, fertility, sleep quality and pregnancybased on physiological signals which are collected automatically while asleep, atnight time. At the point when the user awakens, the bracelet is synchronizedto the mobile application and the algorithm of Ava shows the outcomes. Itsmobile application displays the information about pulse and breathing rates,skin temperature, and heart rate variability ratio which determines physiologicalstress. Sleep is checked by the Ava bracelet and it calculates complete rest time,percentage of light contrasted with deep and REM sleep, including the restpatterns. At the point when the woman is pregnant, the mobile application canscreen her weight and show the development on its graphical user interface. Themobile application monitors and informs its user about the baby’s advancementduring pregnancy.When compared with existing solutions, our proposed wearable-based systemhas clear advantages: first, it automatically measures blood pressure without theneed to maintain a certain position, it does not use inflatable cuffs, and it iscomfortable and easy to wear continuously. Measurement accuracy is ensured bythe design of the device, in which the sensor covers a large surface of the innerwrist. Furthermore, the software server employs machine learning techniques todetect potential signs of preeclampsia [20]. Considering the above, the systemwe propose is original, technologically advanced and accessible both with regardto cost and ease of use.
The platform provides an end-to-end solution for monitoring the onset and pro-gression of hypertension-related conditions tailored for pregnant women. In addi-tion to the custom-developed smart bracelet, it incorporates one of the end-user’ssmart mobile devices. A custom developed application is installed on this device,which is usually an Android or iOS powered smartphone or tablet. It is used tocommunicate with the bracelet using the low power BLE standard, process rawreadings and transmit them to the software server. This fills the gap between thepower-constrained environment of the bracelet and the cloud-deployed softwareserver. The clinician associated with the patient can consult her reading historythrough a web interface. Clinicians can sign up to receive real-time alerts in casethe system detects the onset of hypertension. The present section presents thehardware and software components and details the design decisions behind thesystem architecture shown in Figure 1.
The wearable bracelet device is comprised of innovative biosensors that monitorthe user’s blood pressure. They measure the microfluidic channel deformation yber-Physical Platform for Preeclampsia Detection 5
Fig. 1.
High-level architecture of the platform based on the impedance principle. Their large size covers a sufficient area of thewrist to detect the small changes in pressure as blood pushes against the wall ofthe radial artery. The sensor is connected to an electronic module comprised ofa lock in amplifier, a microcontroller, a current generator, a screen, battery anda USB port.The current generator is connected to the electrodes of the sensor used tomeasure the impedance of the deformable microfluidic channel. The lock in am-plifier uses the sensor voltage signals from which noise is extracted. The signalis converted from analog to digital, after which it is forwarded to the micro-controller. The microcontroller is responsible for sensor data acquisition, dataencryption and wireless transmission to the associated user device. The micro-controller encrypts the waveform information and transmits it to the paireddevice over Bluetooth. The bracelet includes a screen used to pair it with a mo-bile device. The power supply for the integrated electronic modules is providedthrough the use of a battery, which can be recharged using a USB port.
The software components are spread across the bracelet, a mobile applicationwith Android and iOS implementations and the software server. The braceletincludes the required software to record and carry out initial analysis on the pres-sure waveform [20], be paired with a mobile device and send encrypted readingdata to it. The mobile application acts as the bridge between the bracelet andthe software server. Its inclusion reduces the bracelet’s weight, complexity andcost, as it only requires a BLE chip for communication. Furthermore, it keepsend users in control of the recorded data, which can be viewed and managedthrough the mobile application.The bracelet’s sensing layer is comprised of several dedicated sensors thatuse the arterial tonometry technique [12] to calculate hemodynamic parameters.
Marin et al.
Fig. 2.
Mobile application user interface screens
When equipped, the dedicated sensors are positioned along a peripheral artery(usually the radial artery) which they make contact with without flattening itor disturbing blood flow. The waveform of the pulse is captured by the sensorand sent to the microcontroller, which discards corrupt data sequences caused byincorrect placement of the bracelet. Correct waveforms are sent to the paired mo-bile device. The accompanying application determines minimum and maximumvalues based on the Hill Climbing method [23]. The algorithm is implemented atmobile device level to maximize the bracelet’s battery life. The most importantparameters obtained are the systolic and diastolic blood pressure values.Users can monitor blood pressure values using the mobile application orthrough the software server’s web interface. Both offer the possibility of register-ing family members or physicians, giving them access to data. Physicians havethe possibility of receiving real-time alerts at the onset of persistent hyperten-sion. As shown in Figure 2, the application also records heart rate. The user canconfigure the measurement frequency, mark the period of time they reserve forresting and configure how they choose to be notified. The settings are designed toprovide balance between the frequency of measurement, device battery life andavoiding the psychological observer effect. Furthermore, values recorded duringthe resting period are important, as they might signal increased risk [7].The risk of preeclampsia is present when persistent hypertension is detected.This is defined as systolic values over 140 mmHg or diastolic values over 90mmHg recorded over two measurements at least 6 hours apart [5]. In this case,the mobile application can raise an alert to notify the end user to contact their yber-Physical Platform for Preeclampsia Detection 7 physician. All recorded values are stored locally by the mobile application. Oncean Internet connection is available, they are sent to the application server.The server is the central repository where all recorded blood pressure valuesare stored. End-users and physicians access the server through a web applicationthat enforces role-based access to ensure data privacy. The application allows allusers to visualize recorded values, statistical data and alert history. The serveralso carries out the final, but most complex phase of data processing. Physi-cians can enter additional contextual and clinical information into the systemsuch as patient age, weight, height, race, smoking status and cholesterol level.These values are fed to a machine learning classifier - one of the project’s inno-vations on the software side. The classifier was trained using the open data setfrom the Massachusetts University Amherst and National Health and NutritionSurvey and is described in [20]. It provides a finely tuned risk profile for hyper-tensive conditions in pregnancy and identifies those cases that are most proneto preeclampsia, before physiological changes (e.g. proteinuria) can be detected.
The most important challenge regarded the accuracy of the waveform-basedmeasurement in the context of a permanently-equipped bracelet. The researchteam considered placing the device both on the arm, like a regular blood pressurecuff and on the wrist, similar to a watch.The main trade-off concerns the fact that the arm is closer to heart and at thesame level, which increases the likelihood of accurate readings; since technicalevaluation showed that a wrist-mounted bracelet produces accurate readings, itwas preferred. This also has the added advantage of not looking out of place, asthe device resembles a fitness tracker . Additional optimizations were carried outto minimize battery consumption and ensure that normal wear and tear does notaffect reading accuracy or result in leaks from the microfluidic layer. Dividingsoftware components across the bracelet, mobile device and server complementshardware design choices, lowers entry cost and allows additional features relatedto disease identification and early warning to be evaluated and rolled out on theserver side. Hypertensive disorders can appear at any time during pregnancy, while preeclamp-sia is usually reported after 20 weeks. Early detection is paramount to give themother the opportunity to receive proper treatment and minimise risks duringpregnancy and delivery. The platform described is part of the i-bracelet project[10], which aims to contribute towards decreasing hypertension related risks inpregnant women. The current paper is focused on the software components,while hardware aspects are discussed in our previous work [20,18,19]. Further details are subject to patent protection. Marin et al.
When compared with existing approaches, our proposal provides an end-to-end solution for continuous, real-time monitoring that combines an innovative,easy to use piece of hardware integrated with software services that enable datacollection, advanced data analysis to predict future cases of persistent hyperten-sion together with statistical reporting.From the hardware side, we believe the greatest improvement will be to ex-tend the project’s scope in order to cover prevention and management of hyper-tension in the general population. The best way to achieve this is to integrate thebracelet’s technology into well-known wearable devices. In the interim, we aimto integrate recordings with globally used platforms such as Apple HealthKit and Google Fit . On the server side, we have evaluated several machine learningalgorithms and obtained promising results [20,19]. We intend to further improveclassification and increase its predictive power, both in the particular case ofpregnant women and within the general population. Acknowledgement
This work was funded by a grant of the Romanian National Authority for Sci-entific Research and Innovation, CCCDI-UEFISCDI, project number 59/2017,Eurostars Project E10871, i-bracelet- ”Intelligent bracelet for blood pressure mon-itoring and detection of preeclampsia” . References
1. Ava: Ava Bracelet. (2020), accessed on: 10.03.20202. Braunthal, S., Brateanu, A.: Hypertension in pregnancy: Pathophysiology andtreatment. SAGE open medicine (2019)3. Carek, A.M., Conant, J., Joshi, A., Kang, H., Inan, O.T.: Seismowatch: Wearablecuffless blood pressure monitoring using pulse transit time. Proc. ACM Interact.Mob. Wearable Ubiquitous Technol. (3) (Sep 2017)4. Dias, D., Cunha, J.P.: Wearable health devicesvital sign monitoring, systems andtechnologies. Sensors , 2414 (07 2018)5. English, F.A., Kenny, L.C., McCarthy, F.P.: Risk factors and effective managementof preeclampsia. Integrated blood pressure control , 7 (2015)6. Faith Mueni Musyoka, M.M.T., Muketha, G.M.: An Assessment of Suit-able and Affordable Smart armband for preeclampsia Management in An-tenatal Care. International Journal of Information Technology (6), 1–7(Feb 2020). https://doi.org/10.5281/zenodo.3685168, https://doi.org/10.5281/zenodo.3685168
7. Friedman, O., Logan, A.G.: Nocturnal blood pressure profiles among normotensive,controlled hypertensive and refractory hypertensive subjects. Canadian Journal ofCardiology (9), S312–S316 (2009)8. Goldenberg, R.L., McClure, E.M.: It takes a system: magnesium sulfate for preven-tion of eclampsia in a resource-limited community setting. Glob Health Sci Pract. , 340–343 (9 2019) https://developer.apple.com/healthkit/ https://h2care.com/18 (2018), accessed on: 28.01.202010. Home of i-bracelet: http://i-bracelet.eu/ (2020)11. Hsu, Y.P., Young, D.J.: Skin-coupled personal wearable ambulatory pulse wavevelocity monitoring system using microelectromechanical sensors. IEEE SensorsJournal (10), 3490–3497 (2014)12. Kemmotsu, O., Ueda, M., Otsuka, H., Yamamura, T., Winter, D.C., Eckerle, J.S.:Arterial tonometry for noninvasive, continuous blood pressure monitoring duringanesthesia. Anesthesiology (2), 333–340 (1991)13. Kongwattanakul, K., Saksiriwuttho, P., Chaiyarach, S., Thepsuthammarat, K.:Incidence, characteristics, maternal complications, and perinatal outcomes associ-ated with preeclampsia with severe features and HELLP syndrome. Internationaljournal of women’s health , 371 (2018)14. Kumar, S., Nilsen, W.J., Abernethy, A., Atienza, A., Patrick, K., Pavel, M., Riley,W.T., Shar, A., Spring, B., Spruijt-Metz, D., Hedeker, D., Honavar, V., Kravitz,R., Lefebvre, R.C., Mohr, D.C., Murphy, S.A., Quinn, C., Shusterman, V., Swen-deman, D.: Mobile health technology evaluation: The mhealth evidence workshop.American Journal of Preventive Medicine (2), 228 – 236 (2013)15. Lo, J., Mission, J., Caughey, A.: Hypertensive disease of pregnancy and maternalmortality. Current opinion in obstetrics & gynecology , 124–32 (04 2013)16. Majumder, S., Mondal, T., Deen, M.: Wearable sensors for remote health moni-toring. Sensors (01 2017)17. Mammaro, A., Carrara, S., Cavaliere, A., Ermito, S., Dinatale, A., Pappalardo,E.M., Militello, M., Pedata, R.: Hypertensive disorders of pregnancy. Journal ofprenatal medicine (1), 1 (2009)18. Marin, I., Goga, N.: Securing the network for a smart bracelet system. In: 2018 22ndInternational Conference on System Theory, Control and Computing (ICSTCC).pp. 255–260 (Oct 2018)19. Marin, I., Pavaloiu, B., Marian, C., Racovita, V., Goga, N.: Early detection ofpreeclampsia based on a machine learning approach. In: 2019 E-Health and Bio-engineering Conference (EHB). pp. 1–4 (Nov 2019)20. Marin, I., Goga, N.: Hypertension detection based on machine learning. In: Pro-ceedings of the 6th Conference on the Engineering of Computer Based Systems.ECBS 19, Association for Computing Machinery, New York, NY, USA (2019)21. Musyoka, F.M., Thiga, M.M., Muketha, G.M.: A 24-hour ambulatory blood pres-sure monitoring system for preeclampsia management in antenatal care. Informat-ics in Medicine Unlocked , 100199 (2019)22. Omron Healthcare: HeartGuide. https://omronhealthcare.com/products/heartguide-wearable-blood-pressure-monitor-bp8000m/ (2020), accessed on:28.01.202023. Rawat, R., Chandel, S.: Hill climbing techniques for tracking maximum power pointin solar photovoltaic systems - a review. Journal of Sustainable Development andGreen Economics , 90–95 (2013)24. Sibai, B.: Diagnosis, prevention, and management of eclampsia. Obstetrics andgynecology , 402–10 (03 2005)25. Vyata, P., Chauhan, N., Nallathambi, A., Hussein, F.: Assessment of prevalenceof preeclampsia from Dilla region of Ethiopia. BMC Research Notes (12 2015)26. Webster, K., Fishburn, S., Maresh, M., Findlay, S.C., Chappell, L.C.: Diagnosisand management of hypertension in pregnancy: summary of updated nice guidance.BMJ366