- Email: [email protected]
Influence of System Configuration on the Quality of Non-Invasive Fetal Electrocardiography Measurement
Available online at www.sciencedirect.com
ScienceDirect IFAC PapersOnLine 52-27 (2019) 421–426
Influence of System Configuration on the Quality of Non-Invasive Fetal Influence of System Configuration on the Quality of Non-Invasive Fetal Influence of System Configuration on the Quality of Non-Invasive Fetal Influence of System Configuration on the Quality of Non-Invasive Fetal Electrocardiography Measurement Influence of System Configuration on the Quality of Non-Invasive Fetal Electrocardiography Measurement Electrocardiography Measurement Electrocardiography Measurement Electrocardiography Measurement Jindrich Brablik, Radana Kahankova, and Radek Martinek
Jindrich Brablik, Radana Kahankova, and Radek Martinek Jindrich Brablik, Radana Kahankova, and Radek Martinek Jindrich Brablik, Radana and Jindrich Brablik, Cybernetics Radana Kahankova, Kahankova, and Radek Radek Martinek Martinek Department Department of of Cybernetics and and Biomedical Biomedical Engineering, Engineering, Department of Cybernetics and Biomedical Engineering, Faculty of Electrical Engineering and Computer Science, VSB Technical University of Ostrava, Ostrava, 17. 17. listopadu listopadu 2172/15, 2172/15, 708 708 Faculty of Electrical EngineeringDepartment and Computer Science, VSBand - Technical University of of Cybernetics Biomedical Engineering, ofOstrava-Poruba, Cybernetics Biomedical Engineering, Faculty of Electrical EngineeringDepartment and Computer Science, VSBand - Technical University of Ostrava, 17. listopadu 2172/15, 708 00 Czech Republic 00 Ostrava-Poruba, Czech Republic Faculty and Science, University of Faculty of of Electrical Electrical Engineering Engineering and Computer Computer Science, VSB VSB -- Technical Technical University of Ostrava, Ostrava, 17. 17. listopadu listopadu 2172/15, 2172/15, 708 708 00 Ostrava-Poruba, Czech Republic [email protected], [email protected], [email protected] [email protected], [email protected], [email protected] 00 Ostrava-Poruba, Czech Republic 00 Ostrava-Poruba, Czech Republic [email protected], [email protected], [email protected] [email protected], [email protected], [email protected] [email protected] Abstract: This paper introduces introduces aa pilot [email protected], study investigating investigating the of various parameters parameters on Abstract: [email protected], paper study the influence influence of various on the the Abstract: This paper introduces a pilot study investigating the influence of various parameters on the quality of electronic fetal monitoring based on non-invasive fetal electrocardiography. The investigation quality of electronic fetal monitoring based on non-invasive fetal electrocardiography. The investigation Abstract: This paper introduces a pilot study investigating the influence of various parameters on Abstract: This paper introduces a in pilot study investigating the influence of various parameters on the the quality of electronic fetal monitoring based on non-invasive fetal electrocardiography. The investigation was carried out as aa necessary step development of an embedded system for fetal monitoring designed was carried out as necessary step in development of an embedded system for fetal monitoring designed quality of electronic fetal monitoring based on non-invasive fetal electrocardiography. The investigation quality of electronic fetal monitoring based on non-invasive fetal electrocardiography. The investigation was carried out as a necessary step in development of an embedded system for fetal monitoring designed as wearable The included four configurations on as wearable device. The experiments experiments included measurements measurements ofsystem four different different configurations on aa wasaa carried carried out device. as aa necessary necessary step in in development development of an an embedded embeddedof system for fetal fetal monitoring monitoring designed was out as step of for designed as a wearable device. The experiments included measurements of four biosignal different configurations on a subject at 34th week of pregnancy by means of 2.0 generation g.USBamp amplifier from g.tec subject at 34th week of pregnancy by means of 2.0 generation g.USBamp biosignal amplifier from g.tec as a wearable device. The experiments included measurements of four different configurations on as a wearable device.of The experiments included measurements of four biosignal different configurations on aa subject atengineering 34th week pregnancy bystudy means of 2.0in generation g.USBamp amplifier from g.tec medical company. The results the recommendation of the most suitable system medical engineering company. The study results in the recommendation of the most suitable system subject at 34th week of pregnancy by means of 2.0 generation g.USBamp biosignal amplifier from g.tec subject 34thand week of pregnancy means of 2.0ingeneration g.USBamp biosignal amplifier from g.tec medicalatengineering company. Thebystudy results thesignal recommendation of clinical the most suitable system configuration electrode in terms the quality feasibility. configuration and the the company. electrode placement placement inresults terms of of the signal quality and and the the clinical feasibility. medical engineering The study in the recommendation of the most suitable system medical engineering company. The study results in the recommendation of the most suitable system configuration and the electrode placement in terms of the signal quality and the clinical feasibility. configuration and the placement in of the quality and clinical © 2019, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All feasibility. rights reserved. configuration and the electrode electrode placement in terms terms of the signal signal quality and the the clinical feasibility. Keywords: fetal electrocardiography, NI-fECG, electronic fetal monitoring, system configuration Keywords: fetal electrocardiography, NI-fECG, electronic fetal monitoring, system configuration Keywords: fetal electrocardiography, NI-fECG, electronic fetal monitoring, system configuration Keywords: fetal Keywords: fetal fetal electrocardiography, electrocardiography, NI-fECG, NI-fECG, electronic electronic fetal monitoring, monitoring, system system configuration configuration positive profile, which results in aa significant significant economic economic loss. loss. positive profile, which results in 1. 1. INTRODUCTION INTRODUCTION positive profile, which results in a significant economicasloss. Despite all mentioned, Doppler-based EFM continues continues the Despite all mentioned, Doppler-based EFM as the positive profile, which results in a significant economic 1. INTRODUCTION positive profile, which results a significant economicasloss. loss. Despite all mentioned, Doppler-based EFM continues the 1. INTRODUCTION standard of care care throughout theinWorld. World. 1. INTRODUCTION The fetal heart rate monitoring (fHR) in its early form standard of throughout the Despite all mentioned, Doppler-based EFM continues as The fetal heart rate monitoring (fHR) in its early form Despite all mentioned, Doppler-based EFM continues as the the standard of care throughout the World. The fetal heart rate monitoring (fHR) in its early form Fetal ofElectrocardiography Electrocardiography (fECG) is is one one of of the the most most was based on the intermittent observations of fetal heart standard care throughout the World. Fetal (fECG) wasThe based on the intermittent observations of fetal heart fetal heart rate monitoring (fHR) in its early form standard of care throughout the World. The fetalonheart rate monitoring (fHR) in and its early form promising Fetal Electrocardiography is one of the most was based the intermittent observations of technology fetal heart methods in in terms terms(fECG) of replacing replacing conventional sounds. Growing development of the promising methods of conventional Fetal Electrocardiography (fECG) is one of the sounds. Growing development ofobservations the science science and technology was based on the intermittent of fetal heart Fetal Electrocardiography (fECG) is one of the most most was based on the intermittent observations of fetal heart promising methods in terms of replacing conventional sounds. Growing development of the science and technology monitoring methods using CTG (Reinhard et al., 2010). The enabled introduction of the first fetal monitors based on monitoring methods using CTG (Reinhard et al., 2010). The enabled introduction of the first fetal monitors based on promising in terms of replacing conventional sounds. Growing Growing development development of of the the science science and and technology technology promising methods in terms of replacing conventional sounds. monitoring methods using CTG (Reinhard et al., 2010). The enabled introductionin ofthethemid-20th first fetal monitors based on fetal fetal well-being is assessed assessed based on the the electric electric potentials phonocardiography century (Smyth et al., al., well-being is based on potentials phonocardiography in the mid-20th century (Smyth et monitoring methods using CTG (Reinhard et al., 2010). The enabled introduction of the first fetal monitors based on methods using CTG (Reinhard et al., 2010). The enabled introduction themid-20th first fetal monitors based on monitoring fetal well-being isofassessed based on the electric potentials phonocardiography in ofthe century (Smyth et al., sensed by means electrodes placed on the maternal body 1958). However, these devices were not able to differentiate sensed by means of electrodes placed on the maternal body 1958). However, these devices were not able to differentiate fetal well-being is assessed based on the electric potentials phonocardiography in the mid-20th century (Smyth et al., fetal well-being is assessed based on the electric potentials phonocardiography in the mid-20th century (Smyth et al., sensed by means of electrodes on the body 1958). However, these devices were not to differentiate (non-invasive method, NI-fECG)placed or directly directly onmaternal the fetal fetal scalp scalp between the maternal maternal and fetal fetal heartable sounds, so the the (non-invasive method, NI-fECG) or on the between the and heart so sensed by means of placed on maternal body 1958). However, these were not able to sensed byfECG means of electrodes electrodes placed on the the maternal body 1958). However, these devices devices werenot not ablesounds, to differentiate differentiate (non-invasive method, NI-fECG) or directly on the fetal scalp between the maternal and was fetal heart sounds, so time the (invasive (invasive monitoring). The fetal heart rate is calculated automated fHR determination possible at that fECG monitoring). The fetal heart rate is calculated automated fHR determination was not possible at that time (non-invasive method, NI-fECG) or directly on the fetal scalp between the maternal and fetal heart sounds, so the (non-invasive method, NI-fECG) or directly on the fetal scalp between the maternal and fetal heart sounds, so the (invasive fECG monitoring). The fetal heart rate is calculated automated fHR determination was not possible at that time from the detected detected RR intervals intervalsThe andfetal thusheart this method method is able able to to (Carrera et al., al., 2003). the RR and thus this is (Carrera et (invasive fECG monitoring). rate automated fHR determination (invasive fECGvariability monitoring). The fetal rate is is calculated calculated automated fHR2003). determination was was not not possible possible at at that that time time from from the detected RR intervals and thusheart this (Jezewski method is able to (Carrera et with al., 2003). monitor fHR more accurately et al., Along that, in 1953, among the first attempts to fHR variability moreand accurately (Jezewski et al., from the detected RR intervals thus this method is able to (Carrera al., 2003). Alonget with that, in 1953, among the first attempts to monitor from the detected RR intervals and thus this method is able to (Carrera et al., 2003). monitor fHR variability more accurately (Jezewski et al., 2017). Moreover, both mother mother and fetus are are not exposed exposed to Along with in 1953, among first attempts to 2017). monitor fHR that, continuously by the means of fetal fetal both fetus not to monitorMoreover, fHR variability variability more and accurately (Jezewski et al., al., monitor fHR continuously by means of Along with that, in 1953, among the first attempts to monitor fHR more accurately (Jezewski et 2017). Moreover, both mother and fetus are not exposed to Along with that, in 1953, among the first attempts to any kind of radiation. radiation. Uterineandcontractions contractions can be also also monitor fHR continuously by etmeans of Infetal electrocardiography took al., 1953). the any kind of Uterine be 2017). Moreover, both mother fetus are not exposed to electrocardiography took place place (Smyth (Smyth etmeans al., 1953). Infetal the monitor fHR continuously by of 2017). Moreover, boththe mother andcontractions fetus areon notcan exposed to any kind of sensing radiation. Uterine can be also monitor fHR several continuously by focused of fetal monitored by electrical activity the maternal electrocardiography tookresearch place (Smyth etmeans al., 1953). Intopic the monitored coming decade, groups on this by sensing the electrical activity on the maternal any kind of radiation. Uterine contractions can be also coming decade, several research groups focused on this topic electrocardiography took place (Smyth et al., 1953). In the any kind of radiation. Uterine contractions can be also by sensing electrical activity on the maternal electrocardiography tookresearch place (Smyth et al., 1953). the monitored abdomen. This methodthe is known known as electrohysterography electrohysterography and coming decade, several groups focused on thisIntopic consequently introduced improvements that being used This method is as and monitored by the electrical activity on maternal consequently introduced improvements that are areon being used abdomen. coming decade, several research groups focused this topic monitored by sensing sensing the electrical activity on the the maternal abdomen. This method is known as electrohysterography and coming decade, several research groups focused on this topic it is potentially more accurate for uterine activity monitoring consequently introduced improvements that are being used for in current obstetrics, obstetrics, such asused the it is potentially more accurate foras activity monitoring abdomen. This is electrohysterography and for internal internal monitoring monitoring in current such as the consequently introduced improvements that are being abdomen. This method method is known known asuterine electrohysterography and it is potentially more accurate for uterine activity monitoring consequently introduced improvements that are being used than intra-uterine pressure catheter (Jacod et al., 2010). for internal monitoring in current obstetrics, such as the modern type of intrauterine catheter (Williams et al., 1952), than intra-uterine pressure catheter (Jacod et al., 2010). it is potentially more accurate for uterine activity monitoring modern type of intrauterine catheter (Williams et al., 1952), for internal internal monitoring monitoring in in current current obstetrics, obstetrics, such such as as the the it is potentially more accurate for uterine activity monitoring than intra-uterine pressure catheter (Jacod et al., 2010). for Therefore, NI-fECG is capable capable to(Jacod fully et replace replace the modern typeelectrode of intrauterine al., 1952), Therefore, fetal scalp scalp (Huntercatheter et al., al., (Williams 1964), or or et cancellation NI-fECG is to fully the than intra-uterine pressure catheter al., fetal (Hunter et 1964), modern type of catheter (Williams et al., 1952), than intra-uterine pressure catheter (Jacod et replace al., 2010). 2010). Therefore, NI-fECG isby capable toCTG fully the modern typeelectrode ofetintrauterine intrauterine catheter etcancellation al., 1952), conventional monitoring means of (Graatsma et al., fetal scalp electrode (Hunter et al., (Williams 1964), orled cancellation system (Hon al., 1957). These findings to better monitoringis by capable means ofto CTG (Graatsma et al., Therefore, NI-fECG fully replace the system (Hon et al., (Hunter 1957). These findings led to better conventional fetal scalp electrode et al., 1964), or cancellation Therefore, NI-fECG is capable to fully replace the conventional monitoring by means of CTG (Graatsma et al., fetal scalp electrode (Hunter et al., 1964), or cancellation 2010). system (Hon etofal.,correlations 1957). These findings the led measured to better 2010). understanding be-tween conventional monitoring monitoring by by means means of of CTG CTG (Graatsma (Graatsma et et al., al., understanding of correlations be-tween the measured system (Hon et al., 1957). These findings led to better conventional 2010). system (Hon et al., 1957). These findings led to better understanding of (fHR, correlations be-tween the measured Since for for NI-fECG NI-fECG monitoring monitoring the the ECG ECG electrode electrode is is not not biological signals uterine etc.) and Since biological signals (fHR, uterine contractions, contractions, etc.)measured and the the 2010). understanding of correlations be-tween the 2010). understanding of(Hon correlations be-tween the measured Sinceattached for NI-fECG monitoring the ECG electrode is not biological signals (fHR, uterine contractions, etc.) and the directly directly to the fetus, it suffers from high amount of fetal health state et al., 1996). Based on that in 1969, to the monitoring fetus, it suffers from electrode high amount of fetal healthsignals state (Hon et uterine al., 1996). Based on etc.) that in 1969, Since for NI-fECG the ECG is biological (fHR, contractions, and the Sinceattached forand NI-fECG the sensed ECG is not not biological (fHR,et uterine contractions, and the directly attached to the monitoring fetus, it suffers from electrode high amount of fetal healthsignals state (Hon al.,published 1996). Based on etc.) that in 1969, interference artifacts that are being with the signal Huntingford and Pendleton the first classification interference and artifacts that are being sensed withamount the signal Huntingford and (Hon Pendleton published the first classification directly attached to the fetus, it suffers from high of fetal health state et al., 1996). Based on that in 1969, directly attached to the fetus, it suffers from high amount of fetal health state (Hon et al., 1996). Based on that in 1969, interference and artifacts that are being sensed with the signal Huntingford and Pendleton published the first classification of interest. Specifically, Specifically, maternal electrocardiogram (mECG) system of fetal heart rate (Huntingford & Pendleton, 1969). of interest. maternal electrocardiogram (mECG) system of fetal heart rate (Huntingford & Pendleton, 1969). interference and artifacts that are being sensed with the signal Huntingford and Pendleton published the first classification interference and artifacts that are being sensed with the signal Huntingford and Pendleton published & thePendleton, first classification of interest. Specifically, maternal electrocardiogram (mECG) system of fetal heart rate (Huntingford 1969). contains almost the same same frequencies and temporally temporally is Finally inheartlate late 1960s, the the ultrasonic1969). fetal contains almost the frequencies and is of interest. Specifically, maternal electrocardiogram (mECG) system of fetal fetalin rate (Huntingford (Huntingford & Pendleton, Pendleton, Finally 1960s, ultrasonic fetal of interest.concurrent Specifically, maternal electrocardiogram system of rate & 1969). contains almost thewith same frequencies and a temporally is recorded fECG, and it is is not not trivial(mECG) task to to Finally inheartlate 1960s, the ultrasonic fetal Cardiotocography (CTG), a non-invasive method for recorded concurrent with fECG, and it a trivial task contains almost the same frequencies and temporally is Cardiotocography (CTG),1960s, a non-invasive method fetal for contains Finally in late the ultrasonic almost thewith same frequencies andadvanced temporally is recorded concurrent fECG, and it is not a trivial task to Finally in late 1960s, the ultrasonic fetal separate them (Christov et al., 2013). Thus, signal Cardiotocography (CTG), a non-invasive method for simultaneous fHR and uterine contractions monitoring, was separate them (Christov et al., 2013). Thus, advanced signal recorded concurrent with fECG, and it is not a trivial task to simultaneous fHR and uterine contractions monitoring, was Cardiotocography (CTG), a non-invasive method for recorded concurrent with fECG, and it is not a trivial task to separate them (Christovcan et al., Cardiotocography (CTG), non-invasive method for processing methods be 2013). helpfulThus, in advanced increasingsignal the simultaneous fHR and uterine introduced. Subsequently, thisa contractions method was monitoring, accepted by was the processing methods can be helpful in increasing the separate them (Christov et 2013). Thus, advanced signal introduced. Subsequently, this method was accepted by the simultaneous fHR and uterine contractions monitoring, was separate them (Christov et al., al., 2013). Thus, advanced signal processing methods can be helpful in increasing the simultaneous fHR and uterine contractions monitoring, was applicability of NI-fECG introduced. Subsequently, this method was accepted by the medical community, enteredthis themethod deliverywas rooms in 1968 1968 with of NI-fECG processing methods can be helpful in increasing the medical community, entered the delivery rooms in introduced. Subsequently, accepted bywith the applicability processing methods can be helpful in increasing the applicability of NI-fECG introduced. Subsequently, this method was accepted by the medical entered the delivery in 1968 with applicability Nevertheless, contrary to to adults adults ECG ECG research, research, the the the first firstcommunity, commercially available fetal rooms monitor, Hewlettof NI-fECG Nevertheless, contrary the commercially available fetal monitor, Hewlettmedical community, entered the delivery rooms in 1968 with applicability of NI-fECG medical community, entered the delivery rooms in 1968 with Nevertheless, contrary ECG research, the the first8020A, commercially available fetaltoday. monitor, Hewlett- research research community focused to on adults fetal ECG ECG signal processing Packard and remains remains there until until community focused fetal signal processing Packard and there Nevertheless, contrary to adults ECG research, the the first commercially available fetal monitor, Nevertheless, contrary toon adults ECG research, the the first8020A, commercially available fetaltoday. monitor, HewlettHewlett- research community focused on fetal ECG signal processing Packard 8020A, and remains there until today. suffers from lack of open access databases for the evaluation Although CTG isremains the most most commonly used method method of of suffers from lack of open access databases for the evaluation research community focused on fetal ECG signal processing Packard 8020A, and there until today. Although CTG is the commonly used research community focused on fetal ECG signal processing Packard 8020A, and remains there until today. suffers from lack of open access databases for the evaluation of the from algorithms. Moreover, this method has not been been Although CTGmonitoring is the most commonly method of electronic fetal (EFM), there used is number number of of the algorithms. Moreover, this method not suffers lack of access databases for the evaluation electronic fetal monitoring (EFM), there is Although CTG is the most commonly used method of suffers from lack of open open access databases for has the evaluation of the algorithms. Moreover, this method has not been Although CTG is the most commonly used method of standardized in terms of electrode placement, which varies electronic fetal monitoring (EFM), there is number of evidence that decreases its credibility credibility (Euliano et al., 2013 2013 standardized in termsMoreover, of electrode placement, which varies of the algorithms. this method has not been evidence that decreases its (Euliano et al., electronic fetal monitoring (EFM), there is number of of the algorithms. Moreover, this method has not been standardized in terms of electrode placement, which varies electronic fetal monitoring (EFM), there is number of according to in theterms fetal ofposition position (Rooijakkers etwhich al., 2014). 2014). evidence that decreases its credibility (Euliano et al., 2013 and Sartwelle, 2012). According According to Sartwelle Sartwelle (2012), CTG is according to the fetal (Rooijakkers et al., standardized electrode placement, varies and Sartwelle, 2012). to (2012), CTG is evidence that decreases its credibility (Euliano et al., 2013 standardized in terms of electrode placement, which varies according to the fetal position (Rooijakkers et al., 2014). evidence that decreases its credibility (Euliano et al., 2013 Therefore, each offetal the available available databases includes different and Sartwelle, 2012). for According to Sartwelle (2012), is Therefore, unreliable especially the courtroom courtroom use due due to its itsCTG falseeach the includes according to to theof position databases (Rooijakkers et al., al.,different 2014). unreliable especially the use to falseand Sartwelle, 2012). According to (2012), CTG is the position (Rooijakkers et 2014). Therefore, each offetal the available databases includes different and Sartwelle, 2012). for According to Sartwelle Sartwelle (2012), is according unreliable especially for the courtroom use due to itsCTG falseTherefore, each of the available databases includes different unreliable especially for the courtroom use due to its falseTherefore, each of the available databases includes different 2405-8963 2019, IFAC for (International Federation Automatic Control) Hosting by Elsevier Ltd. All rights reserved. unreliable© especially the courtroom use ofdue to its falsePeer review under responsibility of International Federation of Automatic Control. 10.1016/j.ifacol.2019.12.700
Jindrich Brablik et al. / IFAC PapersOnLine 52-27 (2019) 421–426
422
data and it makes an objective evaluation nearly impossible. Some authors (Behar et al., 2014, or Martinek et al., 2016) introduced synthetic signal generators to produce data for their experiments, however, the results obtained using artificial test signals often differ from those performed on signals from clinical practice. Therefore, the real measurements need to be carried out in order to provide the evaluation of the system performance. In this paper, we introduce the practical issues associated with the real measurements of the NI-fECG. We also propose the optimal system configuration to acquire high quality NIfECG recordings. This is a vital condition for the fECG extraction algorithms to perform well. We also provide the evaluation of various measurement deployments. Based on that, we suggest the optimal electrode placement in accordance with the system configuration in terms of its clinical feasibility and the quality of the output signals.
active ground (black) causes significant changes in the recorded signals since it may help in minimizing both the polarization potential and the maternal component. Therefore, it may significantly influence the performance of the extraction algorithms. The optimal number of electrodes may also differ for each extraction algorithm since blind source separation methods (such as independent Component analysis or principle component analysis) performs better with high number of abdominal inputs whereas a multi-lead system using an adaptive algorithm requires low number of abdominal electrodes but at least one thoracic reference electrode (Kahankova et al., 2019)..
2. PRACTICAL ISSUES This chapter introduces the main practical issues associated with the fECG measurement, namely electrode placement and system configuration. 2.1 Electrode Types and Placement Different types of NI-fECG devices use diverse types electrodes and their deployment. Various elelectrodes and the approaches are associated with improving Common-Mode Rejection. There are following types of electrodes used for NI-fECG monitoring: 1.
2.
3.
Sensing electrodes (SE). These electrodes measure the abdominal or thoracic signals, which are used as the inputs to the extraction system. The number of these electrodes differs among different commercial devices or research papers starting from 1 electrode (used for single channel techniques). The maximum of the electrodes used is associated with the clinical feasibility and both financial and computational costs. Common (or reference) electrode (CE). In (HayesGill et al., 2003), the authors recommend the CE to be placed opposing the location of at least 3 sensing electrodes such that the line taken between the CE and each of the SE passes through the womb of the pregnant subject. Active Ground/Ground reference (GND). Some authors (Taylor et al., 2003) recommend to place the GND adjacent to the navel, while others prefer to locate it on the sides of the abdomen (Behar et al., 2019), on the back (Vullings, 2010) or even on the left thigh (Jezewski et al., 2012).
There are significant differences in electrode deployment among the databases, different researchers and also the commercially available devices, as illustrated in Fig. 1. While the positioning of the measurement electrodes does only influence the magnitude or polarity of the signal, the placement of common reference electrode (blue) and the
Fig. 1. Examples of the deployment of the measurement electrodes (abdominal – white, and chest– purple), common reference (blue), and the active ground (black): a) from left: commercially available device MONICA AN24, positioning used in publically available databases ADFECGDB and NIFEADB; b) from left: Vullings (2010) and Taylor et al., 2003. 2.2 System Configuration The measurement system configuration is an important factor influencing the quality of the fetal monitoring. Among the most important parameters one has to keep in mind are: 1.
Ground electrode placement. This electrode is used to provide common potential for the measurement system. It is needed for getting differential voltage and significantly increases common mode rejection by subtracting the same voltages (power line, RF and other superposed signals) showing at active (sensing) and reference electrodes. Ground electrode can be place anywhere on the body. Its position is generally not relevant, because it does not contribute to the measured waveform.
2.
Reference electrode placement. This electrode is connected as a second input (Vin-) to differential amplifier. Its function is suppression of the unwanted signals (muscle and movement artefacts) in order to further improve the Common-Mode Rejection. For fetal monitoring the electrode is usually placed close to other sensing electrodes,
Jindrich Brablik et al. / IFAC PapersOnLine 52-27 (2019) 421–426
which may result in lower signal amplitude, but it dramatically suppresses the motion and muscle artefacts. 2.3 Signal Pre-processing The preprocessing of the recorded signal and the extraction of the fetal component are vital for the electronic fetal monitoring system functioning. The main aim of these steps are to prepare the signal and extract the desired information about the fetal well being, usually assessed by the fetal heart rate. The preprocessing stage includes the basic filtration of the signal according to the desired signal time-frequency properties. The most common preprocessing procedures, also utilized in this paper, are: 1.
Bandpass filtering (cut-off frequencies 0.5 and 100 Hz);
2.
Notch filtering (according to the location Notch filter set to 60 Hz or 50 Hz);
3.
Bipolar signal derivation (subtraction of two signal measured by sensing electrodes).
423
3.1 Measurement System As a measurement system we used 2.0 generation g.USBamp biosignal amplifier (Fig. 2) from g.tec medical engineering company, designed mainly for investigation of brain, heart and muscle activity. System features 16 DCcoupled simultaneously sampled analog input channels with 24-bit resolution, sampling frequency up to 38.4 kHz, ± 250 mV and biosignal pre-amplifier. Channels are clustered into four groups per four channels with separate ground and reference inputs, which allows measurement of up to four subjects at the same time. As a preprocessing, internal digital bandpass and notch filters as well as bipolar derivation can be applied to each measured channel. For advanced timing synchronization with external events, digital triggering inputs and outputs can be used. Input channels are designed to be compatible with both passive and active electrodes. System offers easy configuration, setup and high-speed online data processing for SIMULINK and LabVIEW as well as driver package/API for other programming languages.
The fetal ECG extraction can be performed using a great variety of methods, which can be divided based on the inputs needed for their functioning: 1.
2.
Combined Source (CS) methods. These methods are effective in reducing the noise that can be identified and recorded and thus used as the reference input of the adaptive system. In case of the fECG signal, the signal considered as the noise is the maternal ECG, which can be recorded by means of electrode placed on the maternal thorax. This signal is assumed to contain only maternal component. An adaptive algorithm is able to estimate the maternal component contained in the abdominal electrocardiogram based on the given reference signal. By subtracting this estimated mECG signal, the estimated fECG signal can be obtained. Abdominal Electrodes Sourced (AES) methods. These methods use the inputs measured by the abdominal electrodes only and thus do not require the additional chest channels. The maternal component to be suppressed is estimated from those inputs, usually multiple of them.
Recently published review reveals that clinical tests should be performed in order to create a recommendation for electrode placement according to the stage of pregnancy, fetal position, number of fetuses, and the algorithm used for the extraction. This article will provide an initial investigation in this matter. 3.
MATERIAL AND METHODS
In this section, we describe the measurement system and the deployment used for the experiments.
Fig. 2.Measurement system: generation 2.0 g.USBamp (The MathWorks, 2019). For our research we developed a custom application in LabVIEW development environment (using the provided LabVIEW API) that allows complex configuration of g.USBamp system, data logging to .tdms (Technical Data Management Streaming) files and basic signal processing (peak detection and heart rate calculation). Based on brief experimentation we employed following configuration when conducting measurements described in next section. Measurement with this configuration provided the highest quality signals where noise and movement artefacts were significantly suppressed. Sample rate 600 Sa/s Ground and reference inputs for all groups interconnected to single potential. Bandpass 8th order Butterworth approximation filter with low cutoff frequency fL = 0,5 Hz and high cutoff frequency fH = 60 Hz. Notch 4th order Butterworth approximation filter with fL = 48 Hz and fH = 52 Hz Bipolar derivation between sensing electrodes (Fig. 1 white) and selected reference sensing electrode (Fig. 1 blue).
Jindrich Brablik et al. / IFAC PapersOnLine 52-27 (2019) 421–426
424
3.2 Measurement Deployment To ensure feasibility and reproducibility of the experiments and to cover most of the available electrode placements, as depicted in Fig. 1, we designed the universal measurement deployment, see Fig. 3. The design included total of 16 electrodes that were positioned on the body of the pregnant woman – 14 on the abdomen, one on the right ankle, and one on the left thigh. The switch between the individual configurations was performed by changing the inputs of the system. It is important to note that we omitted some of the electrodes that were redundant (i.e. the recordings included almost identical signals). This lowered the number of electrodes in some of the configurations, such as the one used in Taylor et al., 2003.
Fig. 3. Measurement deployment. Sensing electrodes (white 1 – 11), ground reference (black A, C, F), and common reference electrode (blue D, E, F). Four different configurations of the measurement system were tested: 1.
Jezewski. This configuration was used in the ADFECGDB database (Jezewski et al., 2012). It includes 4 SEs (1, 2, 4, and 7), the GND (C) placed on the left tight and the reference (D) on the bottom of the abdomen.
2.
Behar. This configuration was used in the NIFEADB database (Behar et al., 2019). It includes 4 SEs (1, 3, 4, 5, and 7), the GND (A) placed on the left tight and the reference (D) on the bottom of the abdomen.
3.
Vullings. The configuration introduced in the investigation of Vullings (2010) includes 8 SEs (2, 3, 5, 6, 8, 9, 10, and 11), the GND (B) placed on the left waist, and the reference (F) is placed adjacent to the navel.
4.
Taylor. This configuration of Taylor et al. (2003) includes 12 sensing electrodes (1 – 11); the GND (F) is placed adjacent to the navel and the common reference (E) on the right ankle.
Table 1 summarizes the above mentioned configurations and links them to the design of the general measurement deployment introduced in Fig. 3. Table 1: Configurations tested Configuration Name Jezewski Behar Vullings Taylor
Sensing electrodes 1, 2, 4, 7 1, 3, 4, 5, 7 2, 3, 5, 6, 8 – 11 1 – 11
Ground reference C A B F
Reference D D F E
RESULTS The experiments included measurements on a real subject in 34th week of pregnancy. The total of 14 electrodes were positioned on the abdomen in the way to cover most of the commonly used electrode deployment as shown in Fig. 4.
Fig. 4. Measurement deployment in real subject at 34th week of pregnancy. Figures 5 – 8 show the resulting records of each configuration as listed in the Table 1. In each record, the fetal R peaks are marked. In the cases of the Vullings and Taylor configurations, only 5 signals of the best quality were plotted in figures 4 and 5, respectively. The recordings were shifted by 50 uV in order to enhance the clarity of the readings. Besides the ratio between fetal and maternal component, the shape and orientation of the QRS complexes are among the most important factors for the quality of the maternal ECG suppression. For extraction algorithms, the QRS complexes should be either negative or positive. Contrary, if the QRS complex is biphasic, the algorithms show lower performance. In the most of the signals in the Jezewski configuration, the magnitude of the fetal component is low in comparison with the maternal one (see Fig. 5). However, in the signal recorded by means of SE 4 and 7, the ratio between the fetal and maternal magnitude is low. Additionally, their orientation is the same (positive), and thus, the mECG suppression is feasible. In the Behar configuration (see Fig. 6), the magnitude of the fetal component is comparable with the maternal one, especially in the abdominal signal recorded by means of sensing electrode 4, 5, and 7. Moreover, the placement of the sensing electrodes is convenient for the implementation of the configuration into the clinical practice since it offers a
Jindrich Brablik et al. / IFAC PapersOnLine 52-27 (2019) 421–426
425
space for the placement of the conventional CTG probes if needed. It is also feasible to be incorporated into a wearable device (e.g. using belts or patch system).
Fig. 5. The output signals of the sensing electrodes 1, 2, 4, and 7 (as listed in the Table 1) for the Jezewski configuration.
Fig. 6. The output signals of the sensing electrodes 1 – 5 (as listed in the Table 1) for the Behar configuration.
Fig. 7. The output signals of the sensing electrodes 5, 6, 7, 8, and 4 (as listed in the Table 1) for the Vullings configuration.
Fig. 8. The output signals of the sensing electrodes 2, 3, 5, 9, and 7(as listed in the Table 1) for the Taylor configuration. Table 2: Tested configurations
The quality of the recordings in the Vullings configuration can be considered as the lowest (see Fig. 7). Only 2 out of the 8 SE, i.e. the SE electrodes 5 and 6, are sufficient in terms of the fetal component magnitude. In the others, the fetal component is nearly not notable by the naked eye. The orientation of the maternal QRS complex is positive in the upper SEs (5 – 8), for the rest of them, the orientation is biphasic or mostly negative. The magnitude of the maternal QRS complex in the Taylor configuration is highest among all of the tested configurations (see Fig. 8). The magnitude of the fetal component is significantly lower and the maternal QRS complexes are biphasic. Thus, the fetal ECG extraction would be challenging. The above analysis of each record is summarized by Table 2. The results imply that Behar configuration is the most suitable in terms of the number of SE with a high quality signal, maternal:fetal magnitude ratio (m:f ratio). The only slight negative of this configuration is the morphology of the maternal QRS (mQRS) complex, which is biphasic in some of the signals, which could influence the fetal ECG extraction.
Configuration Name Jezewski Behar Vullings Taylor
Best quality signal SE 3, 4
m:f ratio
mQRS
Low
Positive Negative/ biphasic Positive biphasic
2, 3, 4
Low
5, 6 2, 3, 9, 7
High High
CONCLUSION This article investigated the influence of system configuration on the quality of NI-fECG monitoring. The main factors that were considered and tested were the electrode placement and the hardware system configuration. The results showed that the electrode placement herein denoted as Behar is the most suitable in terms of the signal quality and the clinical feasibility. Moreover, the paper introduced the optimal system configuration to ensure high quality input signals. This was necessary step in order to develop an embedded system for fetal monitoring designed as a wearable device for pregnant subjects. In the future research, it is necessary to test the configurations as well as the performance of the extraction algorithms for the different
426
Jindrich Brablik et al. / IFAC PapersOnLine 52-27 (2019) 421–426
subject, in different stage of pregnancy and different fetal positions. ACKNOWLEDGMENT This article was supported by the Ministry of Education of the Czech Republic (Project No. SP2019/85). This work was supported by the European Regional Development Fund in the Research Centre of Advanced Mechatronic Systems project, project number CZ.02.1.01/0.0/0.0/16 019/0000867 within the Operational Programme Research, Development and Education. This work was supported by the European. Regional Development Fund in A Research Platform focused on Industry 4.0 and Robotics in Ostrava project, CZ.02.1.01/0.0/0.0/17 049/0008425 within the Operational Programme Research, Development and Education. This paper has been elaborated in the framework of the grant programme ”Support for Science and Research in the MoraviaSilesia Region 2018" (RRC/10/2018), financed from the budget of the Moravian-Silesian Region. REFERENCES Behar, J. A., Bonnemains, L., Shulgin, V., Oster, J., Ostras, O., & Lakhno, I. (2019). Noninvasive fetal electrocardiography for the detection of fetal arrhythmias. Prenatal diagnosis, 39(3), 178-187. Behar, J., Andreotti, F., Zaunseder, S., Li, Q., Oster, J., & Clifford, G. D. (2014). An ECG simulator for generating maternal-foetal activity mixtures on abdominal ECG recordings. Physiological measurement, 35(8), 1537. Carrera, J. M. (2003). The technological development of fetal surveillance: a long history. In Controversies in Perinatal Medicine (pp. 22-39). CRC Press. Carrera, J. M. (2003). The technological development of fetal surveillance: a long history. In Controversies in Perinatal Medicine (pp. 22-39). CRC Press. Euliano, T. Y., Nguyen, M. T., Darmanjian, S., McGorray, S. P., Euliano, N., Onkala, A., & Gregg, A. R. (2013). Monitoring uterine activity during labor: a comparison of 3 methods. American journal of obstetrics and gynecology, 208(1), 66-e1. Graatsma, E. M. (2010). Monitoring of fetal heart rate and uterine activity (Doctoral dissertation, Utrecht University). Hayes-Gill, B. R., Barratt, C. W., & Pieri, J. F. (2014). U.S. Patent No. 8,880,140. Washington, DC: U.S. Patent and Trademark Office. Hon, E. H., & Hess, O. W. (1957). Instrumentation of fetal electrocardiography. Science, 125(3247), 553-554. Hon, E. H., & Quilligan, E. J. (1996). The classification of fetal heart rate. Childbirth: The medicalization of obstetrics, 2, 339. Hunter, J. C. A., Braunlin, R. J., Lansford, K. G., & Knoebel, S. B. (1964). U.S. Patent No. 3,120,227. Washington, DC: U.S. Patent and Trademark Office. Huntingford, P. J., & Pendleton, H. J. (1969). The clinical application of cardiotocography. BJOG: An International Journal of Obstetrics & Gynaecology, 76(7), 586-595. Christov, I., Simova, I., & Abacherli, R. (2013, September). Cancellation of the ma-ternal and extraction of the fetal
ECG in noninvasive recordings. In Computing in Cardiology Conference (CinC), 2013 (pp. 153-156). IEEE. Jacod, B. C., Graatsma, E. M., Van Hagen, E., & Visser, G. H. (2010). A validation of electrohysterography for uterine activity monitoring during labour. The Journal of Maternal-Fetal & Neonatal Medicine, 23(1), 17-22. Jezewski, J., Wrobel, J., Matonia, A., Horoba, K., Martinek, R., Kupka, T., & Jezewski, M. (2017). Is abdominal fetal electrocardiography an alternative to doppler ultrasound for FHR variability evaluation?. Frontiers in physiology, 8, 305. Jezewski, J., Matonia, A., Kupka, T., Roj, D., and Czabanski, R. ”Determination of fetal heart rate from abdominal signals: evaluation of beat-to-beat accuracy in relation to the direct fetal electrocardiogram,” Biomedizinische Technik/Biomed. Eng., vol. 57, no. 5, pp. 383- 394, 2012. Kahankova, R., Martinek, R., Jaros, R., Behbehani, K., Matonia, A., Jezewski, M., & Behar, J. A. (2019). A Review of Signal Processing Techniques for NonInvasive Fetal Electrocardiography. IEEE Reviews in Biomedical Engineering. Martinek, R., Kelnar, M., Koudelka, P., Vanus, J., Bilik, P., Janku, P., ... & Zidek, J. (2016). A novel LabVIEWbased multi-channel non-invasive abdominal maternalfetal electrocardiogram signal generator. Physiological measurement, 37(2), 238. Reinhard, J., Hayes-Gill, B. R., Yi, Q., Hatzmann, H., & Schiermeier, S. (2010). Comparison of non-invasive fetal electrocardiogram to Doppler cardiotocogram dur-ing the 1st stage of labor. Journal of perinatal medicine, 38(2), 179-185. Rooijakkers, M. J., Song, S., Rabotti, C., Oei, S. G., Bergmans, J. W., Cantatore, E., & Mischi, M. (2014). Influence of electrode placement on signal quality for ambulatory pregnancy monitoring. Computational and mathematical methods in medicine, 2014. Sartwelle, T. P. (2012). Electronic fetal monitoring: a bridge too far. Journal of Legal Medicine, 33(3), 313-379. Smyth, C. N., & Farrow, J. L. (1958). Present place in obstetrics for foetal phonocardiography and electrocardiography. British Medical Journal, 2(5103), 1005. Smyth, C. N. (1953). Experimental electrocardiography of the foetus. The Lancet, 261(6771), 1124-1126. Taylor, M. J., Smith, M. J., Thomas, M., Green, A. R., Cheng, F., Oseku‐Afful, S., ... & Gardiner, H. M. (2003). Non‐invasive fetal electrocardiography in singleton and multiple pregnancies. BJOG: An International Journal of Obstetrics & Gynaecology, 110(7), 668-678. The MathWorks, Inc. G.tec Support from Data Acquisition Toolbox. MathWorks [online]. 2019. Available at: https://www.mathworks.com/hardware-support/gtec.html Vullings, R. (2010). Non-invasive fetal electrocardiogram: analysis and interpretation. Eindhoven: Technische Universteit Eindhoven.—2010.—225 p. Williams, E. A., & Stallworthy, J. A. (1952). A simple method of internal tocography. The Lancet, 259(6703), 330-332.
ScienceDirect IFAC PapersOnLine 52-27 (2019) 421–426
Influence of System Configuration on the Quality of Non-Invasive Fetal Influence of System Configuration on the Quality of Non-Invasive Fetal Influence of System Configuration on the Quality of Non-Invasive Fetal Influence of System Configuration on the Quality of Non-Invasive Fetal Electrocardiography Measurement Influence of System Configuration on the Quality of Non-Invasive Fetal Electrocardiography Measurement Electrocardiography Measurement Electrocardiography Measurement Electrocardiography Measurement Jindrich Brablik, Radana Kahankova, and Radek Martinek
Jindrich Brablik, Radana Kahankova, and Radek Martinek Jindrich Brablik, Radana Kahankova, and Radek Martinek Jindrich Brablik, Radana and Jindrich Brablik, Cybernetics Radana Kahankova, Kahankova, and Radek Radek Martinek Martinek Department Department of of Cybernetics and and Biomedical Biomedical Engineering, Engineering, Department of Cybernetics and Biomedical Engineering, Faculty of Electrical Engineering and Computer Science, VSB Technical University of Ostrava, Ostrava, 17. 17. listopadu listopadu 2172/15, 2172/15, 708 708 Faculty of Electrical EngineeringDepartment and Computer Science, VSBand - Technical University of of Cybernetics Biomedical Engineering, ofOstrava-Poruba, Cybernetics Biomedical Engineering, Faculty of Electrical EngineeringDepartment and Computer Science, VSBand - Technical University of Ostrava, 17. listopadu 2172/15, 708 00 Czech Republic 00 Ostrava-Poruba, Czech Republic Faculty and Science, University of Faculty of of Electrical Electrical Engineering Engineering and Computer Computer Science, VSB VSB -- Technical Technical University of Ostrava, Ostrava, 17. 17. listopadu listopadu 2172/15, 2172/15, 708 708 00 Ostrava-Poruba, Czech Republic [email protected], [email protected], [email protected] [email protected], [email protected], [email protected] 00 Ostrava-Poruba, Czech Republic 00 Ostrava-Poruba, Czech Republic [email protected], [email protected], [email protected] [email protected], [email protected], [email protected] [email protected] Abstract: This paper introduces introduces aa pilot [email protected], study investigating investigating the of various parameters parameters on Abstract: [email protected], paper study the influence influence of various on the the Abstract: This paper introduces a pilot study investigating the influence of various parameters on the quality of electronic fetal monitoring based on non-invasive fetal electrocardiography. The investigation quality of electronic fetal monitoring based on non-invasive fetal electrocardiography. The investigation Abstract: This paper introduces a pilot study investigating the influence of various parameters on Abstract: This paper introduces a in pilot study investigating the influence of various parameters on the the quality of electronic fetal monitoring based on non-invasive fetal electrocardiography. The investigation was carried out as aa necessary step development of an embedded system for fetal monitoring designed was carried out as necessary step in development of an embedded system for fetal monitoring designed quality of electronic fetal monitoring based on non-invasive fetal electrocardiography. The investigation quality of electronic fetal monitoring based on non-invasive fetal electrocardiography. The investigation was carried out as a necessary step in development of an embedded system for fetal monitoring designed as wearable The included four configurations on as wearable device. The experiments experiments included measurements measurements ofsystem four different different configurations on aa wasaa carried carried out device. as aa necessary necessary step in in development development of an an embedded embeddedof system for fetal fetal monitoring monitoring designed was out as step of for designed as a wearable device. The experiments included measurements of four biosignal different configurations on a subject at 34th week of pregnancy by means of 2.0 generation g.USBamp amplifier from g.tec subject at 34th week of pregnancy by means of 2.0 generation g.USBamp biosignal amplifier from g.tec as a wearable device. The experiments included measurements of four different configurations on as a wearable device.of The experiments included measurements of four biosignal different configurations on aa subject atengineering 34th week pregnancy bystudy means of 2.0in generation g.USBamp amplifier from g.tec medical company. The results the recommendation of the most suitable system medical engineering company. The study results in the recommendation of the most suitable system subject at 34th week of pregnancy by means of 2.0 generation g.USBamp biosignal amplifier from g.tec subject 34thand week of pregnancy means of 2.0ingeneration g.USBamp biosignal amplifier from g.tec medicalatengineering company. Thebystudy results thesignal recommendation of clinical the most suitable system configuration electrode in terms the quality feasibility. configuration and the the company. electrode placement placement inresults terms of of the signal quality and and the the clinical feasibility. medical engineering The study in the recommendation of the most suitable system medical engineering company. The study results in the recommendation of the most suitable system configuration and the electrode placement in terms of the signal quality and the clinical feasibility. configuration and the placement in of the quality and clinical © 2019, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All feasibility. rights reserved. configuration and the electrode electrode placement in terms terms of the signal signal quality and the the clinical feasibility. Keywords: fetal electrocardiography, NI-fECG, electronic fetal monitoring, system configuration Keywords: fetal electrocardiography, NI-fECG, electronic fetal monitoring, system configuration Keywords: fetal electrocardiography, NI-fECG, electronic fetal monitoring, system configuration Keywords: fetal Keywords: fetal fetal electrocardiography, electrocardiography, NI-fECG, NI-fECG, electronic electronic fetal monitoring, monitoring, system system configuration configuration positive profile, which results in aa significant significant economic economic loss. loss. positive profile, which results in 1. 1. INTRODUCTION INTRODUCTION positive profile, which results in a significant economicasloss. Despite all mentioned, Doppler-based EFM continues continues the Despite all mentioned, Doppler-based EFM as the positive profile, which results in a significant economic 1. INTRODUCTION positive profile, which results a significant economicasloss. loss. Despite all mentioned, Doppler-based EFM continues the 1. INTRODUCTION standard of care care throughout theinWorld. World. 1. INTRODUCTION The fetal heart rate monitoring (fHR) in its early form standard of throughout the Despite all mentioned, Doppler-based EFM continues as The fetal heart rate monitoring (fHR) in its early form Despite all mentioned, Doppler-based EFM continues as the the standard of care throughout the World. The fetal heart rate monitoring (fHR) in its early form Fetal ofElectrocardiography Electrocardiography (fECG) is is one one of of the the most most was based on the intermittent observations of fetal heart standard care throughout the World. Fetal (fECG) wasThe based on the intermittent observations of fetal heart fetal heart rate monitoring (fHR) in its early form standard of care throughout the World. The fetalonheart rate monitoring (fHR) in and its early form promising Fetal Electrocardiography is one of the most was based the intermittent observations of technology fetal heart methods in in terms terms(fECG) of replacing replacing conventional sounds. Growing development of the promising methods of conventional Fetal Electrocardiography (fECG) is one of the sounds. Growing development ofobservations the science science and technology was based on the intermittent of fetal heart Fetal Electrocardiography (fECG) is one of the most most was based on the intermittent observations of fetal heart promising methods in terms of replacing conventional sounds. Growing development of the science and technology monitoring methods using CTG (Reinhard et al., 2010). The enabled introduction of the first fetal monitors based on monitoring methods using CTG (Reinhard et al., 2010). The enabled introduction of the first fetal monitors based on promising in terms of replacing conventional sounds. Growing Growing development development of of the the science science and and technology technology promising methods in terms of replacing conventional sounds. monitoring methods using CTG (Reinhard et al., 2010). The enabled introductionin ofthethemid-20th first fetal monitors based on fetal fetal well-being is assessed assessed based on the the electric electric potentials phonocardiography century (Smyth et al., al., well-being is based on potentials phonocardiography in the mid-20th century (Smyth et monitoring methods using CTG (Reinhard et al., 2010). The enabled introduction of the first fetal monitors based on methods using CTG (Reinhard et al., 2010). The enabled introduction themid-20th first fetal monitors based on monitoring fetal well-being isofassessed based on the electric potentials phonocardiography in ofthe century (Smyth et al., sensed by means electrodes placed on the maternal body 1958). However, these devices were not able to differentiate sensed by means of electrodes placed on the maternal body 1958). However, these devices were not able to differentiate fetal well-being is assessed based on the electric potentials phonocardiography in the mid-20th century (Smyth et al., fetal well-being is assessed based on the electric potentials phonocardiography in the mid-20th century (Smyth et al., sensed by means of electrodes on the body 1958). However, these devices were not to differentiate (non-invasive method, NI-fECG)placed or directly directly onmaternal the fetal fetal scalp scalp between the maternal maternal and fetal fetal heartable sounds, so the the (non-invasive method, NI-fECG) or on the between the and heart so sensed by means of placed on maternal body 1958). However, these were not able to sensed byfECG means of electrodes electrodes placed on the the maternal body 1958). However, these devices devices werenot not ablesounds, to differentiate differentiate (non-invasive method, NI-fECG) or directly on the fetal scalp between the maternal and was fetal heart sounds, so time the (invasive (invasive monitoring). The fetal heart rate is calculated automated fHR determination possible at that fECG monitoring). The fetal heart rate is calculated automated fHR determination was not possible at that time (non-invasive method, NI-fECG) or directly on the fetal scalp between the maternal and fetal heart sounds, so the (non-invasive method, NI-fECG) or directly on the fetal scalp between the maternal and fetal heart sounds, so the (invasive fECG monitoring). The fetal heart rate is calculated automated fHR determination was not possible at that time from the detected detected RR intervals intervalsThe andfetal thusheart this method method is able able to to (Carrera et al., al., 2003). the RR and thus this is (Carrera et (invasive fECG monitoring). rate automated fHR determination (invasive fECGvariability monitoring). The fetal rate is is calculated calculated automated fHR2003). determination was was not not possible possible at at that that time time from from the detected RR intervals and thusheart this (Jezewski method is able to (Carrera et with al., 2003). monitor fHR more accurately et al., Along that, in 1953, among the first attempts to fHR variability moreand accurately (Jezewski et al., from the detected RR intervals thus this method is able to (Carrera al., 2003). Alonget with that, in 1953, among the first attempts to monitor from the detected RR intervals and thus this method is able to (Carrera et al., 2003). monitor fHR variability more accurately (Jezewski et al., 2017). Moreover, both mother mother and fetus are are not exposed exposed to Along with in 1953, among first attempts to 2017). monitor fHR that, continuously by the means of fetal fetal both fetus not to monitorMoreover, fHR variability variability more and accurately (Jezewski et al., al., monitor fHR continuously by means of Along with that, in 1953, among the first attempts to monitor fHR more accurately (Jezewski et 2017). Moreover, both mother and fetus are not exposed to Along with that, in 1953, among the first attempts to any kind of radiation. radiation. Uterineandcontractions contractions can be also also monitor fHR continuously by etmeans of Infetal electrocardiography took al., 1953). the any kind of Uterine be 2017). Moreover, both mother fetus are not exposed to electrocardiography took place place (Smyth (Smyth etmeans al., 1953). Infetal the monitor fHR continuously by of 2017). Moreover, boththe mother andcontractions fetus areon notcan exposed to any kind of sensing radiation. Uterine can be also monitor fHR several continuously by focused of fetal monitored by electrical activity the maternal electrocardiography tookresearch place (Smyth etmeans al., 1953). Intopic the monitored coming decade, groups on this by sensing the electrical activity on the maternal any kind of radiation. Uterine contractions can be also coming decade, several research groups focused on this topic electrocardiography took place (Smyth et al., 1953). In the any kind of radiation. Uterine contractions can be also by sensing electrical activity on the maternal electrocardiography tookresearch place (Smyth et al., 1953). the monitored abdomen. This methodthe is known known as electrohysterography electrohysterography and coming decade, several groups focused on thisIntopic consequently introduced improvements that being used This method is as and monitored by the electrical activity on maternal consequently introduced improvements that are areon being used abdomen. coming decade, several research groups focused this topic monitored by sensing sensing the electrical activity on the the maternal abdomen. This method is known as electrohysterography and coming decade, several research groups focused on this topic it is potentially more accurate for uterine activity monitoring consequently introduced improvements that are being used for in current obstetrics, obstetrics, such asused the it is potentially more accurate foras activity monitoring abdomen. This is electrohysterography and for internal internal monitoring monitoring in current such as the consequently introduced improvements that are being abdomen. This method method is known known asuterine electrohysterography and it is potentially more accurate for uterine activity monitoring consequently introduced improvements that are being used than intra-uterine pressure catheter (Jacod et al., 2010). for internal monitoring in current obstetrics, such as the modern type of intrauterine catheter (Williams et al., 1952), than intra-uterine pressure catheter (Jacod et al., 2010). it is potentially more accurate for uterine activity monitoring modern type of intrauterine catheter (Williams et al., 1952), for internal internal monitoring monitoring in in current current obstetrics, obstetrics, such such as as the the it is potentially more accurate for uterine activity monitoring than intra-uterine pressure catheter (Jacod et al., 2010). for Therefore, NI-fECG is capable capable to(Jacod fully et replace replace the modern typeelectrode of intrauterine al., 1952), Therefore, fetal scalp scalp (Huntercatheter et al., al., (Williams 1964), or or et cancellation NI-fECG is to fully the than intra-uterine pressure catheter al., fetal (Hunter et 1964), modern type of catheter (Williams et al., 1952), than intra-uterine pressure catheter (Jacod et replace al., 2010). 2010). Therefore, NI-fECG isby capable toCTG fully the modern typeelectrode ofetintrauterine intrauterine catheter etcancellation al., 1952), conventional monitoring means of (Graatsma et al., fetal scalp electrode (Hunter et al., (Williams 1964), orled cancellation system (Hon al., 1957). These findings to better monitoringis by capable means ofto CTG (Graatsma et al., Therefore, NI-fECG fully replace the system (Hon et al., (Hunter 1957). These findings led to better conventional fetal scalp electrode et al., 1964), or cancellation Therefore, NI-fECG is capable to fully replace the conventional monitoring by means of CTG (Graatsma et al., fetal scalp electrode (Hunter et al., 1964), or cancellation 2010). system (Hon etofal.,correlations 1957). These findings the led measured to better 2010). understanding be-tween conventional monitoring monitoring by by means means of of CTG CTG (Graatsma (Graatsma et et al., al., understanding of correlations be-tween the measured system (Hon et al., 1957). These findings led to better conventional 2010). system (Hon et al., 1957). These findings led to better understanding of (fHR, correlations be-tween the measured Since for for NI-fECG NI-fECG monitoring monitoring the the ECG ECG electrode electrode is is not not biological signals uterine etc.) and Since biological signals (fHR, uterine contractions, contractions, etc.)measured and the the 2010). understanding of correlations be-tween the 2010). understanding of(Hon correlations be-tween the measured Sinceattached for NI-fECG monitoring the ECG electrode is not biological signals (fHR, uterine contractions, etc.) and the directly directly to the fetus, it suffers from high amount of fetal health state et al., 1996). Based on that in 1969, to the monitoring fetus, it suffers from electrode high amount of fetal healthsignals state (Hon et uterine al., 1996). Based on etc.) that in 1969, Since for NI-fECG the ECG is biological (fHR, contractions, and the Sinceattached forand NI-fECG the sensed ECG is not not biological (fHR,et uterine contractions, and the directly attached to the monitoring fetus, it suffers from electrode high amount of fetal healthsignals state (Hon al.,published 1996). Based on etc.) that in 1969, interference artifacts that are being with the signal Huntingford and Pendleton the first classification interference and artifacts that are being sensed withamount the signal Huntingford and (Hon Pendleton published the first classification directly attached to the fetus, it suffers from high of fetal health state et al., 1996). Based on that in 1969, directly attached to the fetus, it suffers from high amount of fetal health state (Hon et al., 1996). Based on that in 1969, interference and artifacts that are being sensed with the signal Huntingford and Pendleton published the first classification of interest. Specifically, Specifically, maternal electrocardiogram (mECG) system of fetal heart rate (Huntingford & Pendleton, 1969). of interest. maternal electrocardiogram (mECG) system of fetal heart rate (Huntingford & Pendleton, 1969). interference and artifacts that are being sensed with the signal Huntingford and Pendleton published the first classification interference and artifacts that are being sensed with the signal Huntingford and Pendleton published & thePendleton, first classification of interest. Specifically, maternal electrocardiogram (mECG) system of fetal heart rate (Huntingford 1969). contains almost the same same frequencies and temporally temporally is Finally inheartlate late 1960s, the the ultrasonic1969). fetal contains almost the frequencies and is of interest. Specifically, maternal electrocardiogram (mECG) system of fetal fetalin rate (Huntingford (Huntingford & Pendleton, Pendleton, Finally 1960s, ultrasonic fetal of interest.concurrent Specifically, maternal electrocardiogram system of rate & 1969). contains almost thewith same frequencies and a temporally is recorded fECG, and it is is not not trivial(mECG) task to to Finally inheartlate 1960s, the ultrasonic fetal Cardiotocography (CTG), a non-invasive method for recorded concurrent with fECG, and it a trivial task contains almost the same frequencies and temporally is Cardiotocography (CTG),1960s, a non-invasive method fetal for contains Finally in late the ultrasonic almost thewith same frequencies andadvanced temporally is recorded concurrent fECG, and it is not a trivial task to Finally in late 1960s, the ultrasonic fetal separate them (Christov et al., 2013). Thus, signal Cardiotocography (CTG), a non-invasive method for simultaneous fHR and uterine contractions monitoring, was separate them (Christov et al., 2013). Thus, advanced signal recorded concurrent with fECG, and it is not a trivial task to simultaneous fHR and uterine contractions monitoring, was Cardiotocography (CTG), a non-invasive method for recorded concurrent with fECG, and it is not a trivial task to separate them (Christovcan et al., Cardiotocography (CTG), non-invasive method for processing methods be 2013). helpfulThus, in advanced increasingsignal the simultaneous fHR and uterine introduced. Subsequently, thisa contractions method was monitoring, accepted by was the processing methods can be helpful in increasing the separate them (Christov et 2013). Thus, advanced signal introduced. Subsequently, this method was accepted by the simultaneous fHR and uterine contractions monitoring, was separate them (Christov et al., al., 2013). Thus, advanced signal processing methods can be helpful in increasing the simultaneous fHR and uterine contractions monitoring, was applicability of NI-fECG introduced. Subsequently, this method was accepted by the medical community, enteredthis themethod deliverywas rooms in 1968 1968 with of NI-fECG processing methods can be helpful in increasing the medical community, entered the delivery rooms in introduced. Subsequently, accepted bywith the applicability processing methods can be helpful in increasing the applicability of NI-fECG introduced. Subsequently, this method was accepted by the medical entered the delivery in 1968 with applicability Nevertheless, contrary to to adults adults ECG ECG research, research, the the the first firstcommunity, commercially available fetal rooms monitor, Hewlettof NI-fECG Nevertheless, contrary the commercially available fetal monitor, Hewlettmedical community, entered the delivery rooms in 1968 with applicability of NI-fECG medical community, entered the delivery rooms in 1968 with Nevertheless, contrary ECG research, the the first8020A, commercially available fetaltoday. monitor, Hewlett- research research community focused to on adults fetal ECG ECG signal processing Packard and remains remains there until until community focused fetal signal processing Packard and there Nevertheless, contrary to adults ECG research, the the first commercially available fetal monitor, Nevertheless, contrary toon adults ECG research, the the first8020A, commercially available fetaltoday. monitor, HewlettHewlett- research community focused on fetal ECG signal processing Packard 8020A, and remains there until today. suffers from lack of open access databases for the evaluation Although CTG isremains the most most commonly used method method of of suffers from lack of open access databases for the evaluation research community focused on fetal ECG signal processing Packard 8020A, and there until today. Although CTG is the commonly used research community focused on fetal ECG signal processing Packard 8020A, and remains there until today. suffers from lack of open access databases for the evaluation of the from algorithms. Moreover, this method has not been been Although CTGmonitoring is the most commonly method of electronic fetal (EFM), there used is number number of of the algorithms. Moreover, this method not suffers lack of access databases for the evaluation electronic fetal monitoring (EFM), there is Although CTG is the most commonly used method of suffers from lack of open open access databases for has the evaluation of the algorithms. Moreover, this method has not been Although CTG is the most commonly used method of standardized in terms of electrode placement, which varies electronic fetal monitoring (EFM), there is number of evidence that decreases its credibility credibility (Euliano et al., 2013 2013 standardized in termsMoreover, of electrode placement, which varies of the algorithms. this method has not been evidence that decreases its (Euliano et al., electronic fetal monitoring (EFM), there is number of of the algorithms. Moreover, this method has not been standardized in terms of electrode placement, which varies electronic fetal monitoring (EFM), there is number of according to in theterms fetal ofposition position (Rooijakkers etwhich al., 2014). 2014). evidence that decreases its credibility (Euliano et al., 2013 and Sartwelle, 2012). According According to Sartwelle Sartwelle (2012), CTG is according to the fetal (Rooijakkers et al., standardized electrode placement, varies and Sartwelle, 2012). to (2012), CTG is evidence that decreases its credibility (Euliano et al., 2013 standardized in terms of electrode placement, which varies according to the fetal position (Rooijakkers et al., 2014). evidence that decreases its credibility (Euliano et al., 2013 Therefore, each offetal the available available databases includes different and Sartwelle, 2012). for According to Sartwelle (2012), is Therefore, unreliable especially the courtroom courtroom use due due to its itsCTG falseeach the includes according to to theof position databases (Rooijakkers et al., al.,different 2014). unreliable especially the use to falseand Sartwelle, 2012). According to (2012), CTG is the position (Rooijakkers et 2014). Therefore, each offetal the available databases includes different and Sartwelle, 2012). for According to Sartwelle Sartwelle (2012), is according unreliable especially for the courtroom use due to itsCTG falseTherefore, each of the available databases includes different unreliable especially for the courtroom use due to its falseTherefore, each of the available databases includes different 2405-8963 2019, IFAC for (International Federation Automatic Control) Hosting by Elsevier Ltd. All rights reserved. unreliable© especially the courtroom use ofdue to its falsePeer review under responsibility of International Federation of Automatic Control. 10.1016/j.ifacol.2019.12.700
Jindrich Brablik et al. / IFAC PapersOnLine 52-27 (2019) 421–426
422
data and it makes an objective evaluation nearly impossible. Some authors (Behar et al., 2014, or Martinek et al., 2016) introduced synthetic signal generators to produce data for their experiments, however, the results obtained using artificial test signals often differ from those performed on signals from clinical practice. Therefore, the real measurements need to be carried out in order to provide the evaluation of the system performance. In this paper, we introduce the practical issues associated with the real measurements of the NI-fECG. We also propose the optimal system configuration to acquire high quality NIfECG recordings. This is a vital condition for the fECG extraction algorithms to perform well. We also provide the evaluation of various measurement deployments. Based on that, we suggest the optimal electrode placement in accordance with the system configuration in terms of its clinical feasibility and the quality of the output signals.
active ground (black) causes significant changes in the recorded signals since it may help in minimizing both the polarization potential and the maternal component. Therefore, it may significantly influence the performance of the extraction algorithms. The optimal number of electrodes may also differ for each extraction algorithm since blind source separation methods (such as independent Component analysis or principle component analysis) performs better with high number of abdominal inputs whereas a multi-lead system using an adaptive algorithm requires low number of abdominal electrodes but at least one thoracic reference electrode (Kahankova et al., 2019)..
2. PRACTICAL ISSUES This chapter introduces the main practical issues associated with the fECG measurement, namely electrode placement and system configuration. 2.1 Electrode Types and Placement Different types of NI-fECG devices use diverse types electrodes and their deployment. Various elelectrodes and the approaches are associated with improving Common-Mode Rejection. There are following types of electrodes used for NI-fECG monitoring: 1.
2.
3.
Sensing electrodes (SE). These electrodes measure the abdominal or thoracic signals, which are used as the inputs to the extraction system. The number of these electrodes differs among different commercial devices or research papers starting from 1 electrode (used for single channel techniques). The maximum of the electrodes used is associated with the clinical feasibility and both financial and computational costs. Common (or reference) electrode (CE). In (HayesGill et al., 2003), the authors recommend the CE to be placed opposing the location of at least 3 sensing electrodes such that the line taken between the CE and each of the SE passes through the womb of the pregnant subject. Active Ground/Ground reference (GND). Some authors (Taylor et al., 2003) recommend to place the GND adjacent to the navel, while others prefer to locate it on the sides of the abdomen (Behar et al., 2019), on the back (Vullings, 2010) or even on the left thigh (Jezewski et al., 2012).
There are significant differences in electrode deployment among the databases, different researchers and also the commercially available devices, as illustrated in Fig. 1. While the positioning of the measurement electrodes does only influence the magnitude or polarity of the signal, the placement of common reference electrode (blue) and the
Fig. 1. Examples of the deployment of the measurement electrodes (abdominal – white, and chest– purple), common reference (blue), and the active ground (black): a) from left: commercially available device MONICA AN24, positioning used in publically available databases ADFECGDB and NIFEADB; b) from left: Vullings (2010) and Taylor et al., 2003. 2.2 System Configuration The measurement system configuration is an important factor influencing the quality of the fetal monitoring. Among the most important parameters one has to keep in mind are: 1.
Ground electrode placement. This electrode is used to provide common potential for the measurement system. It is needed for getting differential voltage and significantly increases common mode rejection by subtracting the same voltages (power line, RF and other superposed signals) showing at active (sensing) and reference electrodes. Ground electrode can be place anywhere on the body. Its position is generally not relevant, because it does not contribute to the measured waveform.
2.
Reference electrode placement. This electrode is connected as a second input (Vin-) to differential amplifier. Its function is suppression of the unwanted signals (muscle and movement artefacts) in order to further improve the Common-Mode Rejection. For fetal monitoring the electrode is usually placed close to other sensing electrodes,
Jindrich Brablik et al. / IFAC PapersOnLine 52-27 (2019) 421–426
which may result in lower signal amplitude, but it dramatically suppresses the motion and muscle artefacts. 2.3 Signal Pre-processing The preprocessing of the recorded signal and the extraction of the fetal component are vital for the electronic fetal monitoring system functioning. The main aim of these steps are to prepare the signal and extract the desired information about the fetal well being, usually assessed by the fetal heart rate. The preprocessing stage includes the basic filtration of the signal according to the desired signal time-frequency properties. The most common preprocessing procedures, also utilized in this paper, are: 1.
Bandpass filtering (cut-off frequencies 0.5 and 100 Hz);
2.
Notch filtering (according to the location Notch filter set to 60 Hz or 50 Hz);
3.
Bipolar signal derivation (subtraction of two signal measured by sensing electrodes).
423
3.1 Measurement System As a measurement system we used 2.0 generation g.USBamp biosignal amplifier (Fig. 2) from g.tec medical engineering company, designed mainly for investigation of brain, heart and muscle activity. System features 16 DCcoupled simultaneously sampled analog input channels with 24-bit resolution, sampling frequency up to 38.4 kHz, ± 250 mV and biosignal pre-amplifier. Channels are clustered into four groups per four channels with separate ground and reference inputs, which allows measurement of up to four subjects at the same time. As a preprocessing, internal digital bandpass and notch filters as well as bipolar derivation can be applied to each measured channel. For advanced timing synchronization with external events, digital triggering inputs and outputs can be used. Input channels are designed to be compatible with both passive and active electrodes. System offers easy configuration, setup and high-speed online data processing for SIMULINK and LabVIEW as well as driver package/API for other programming languages.
The fetal ECG extraction can be performed using a great variety of methods, which can be divided based on the inputs needed for their functioning: 1.
2.
Combined Source (CS) methods. These methods are effective in reducing the noise that can be identified and recorded and thus used as the reference input of the adaptive system. In case of the fECG signal, the signal considered as the noise is the maternal ECG, which can be recorded by means of electrode placed on the maternal thorax. This signal is assumed to contain only maternal component. An adaptive algorithm is able to estimate the maternal component contained in the abdominal electrocardiogram based on the given reference signal. By subtracting this estimated mECG signal, the estimated fECG signal can be obtained. Abdominal Electrodes Sourced (AES) methods. These methods use the inputs measured by the abdominal electrodes only and thus do not require the additional chest channels. The maternal component to be suppressed is estimated from those inputs, usually multiple of them.
Recently published review reveals that clinical tests should be performed in order to create a recommendation for electrode placement according to the stage of pregnancy, fetal position, number of fetuses, and the algorithm used for the extraction. This article will provide an initial investigation in this matter. 3.
MATERIAL AND METHODS
In this section, we describe the measurement system and the deployment used for the experiments.
Fig. 2.Measurement system: generation 2.0 g.USBamp (The MathWorks, 2019). For our research we developed a custom application in LabVIEW development environment (using the provided LabVIEW API) that allows complex configuration of g.USBamp system, data logging to .tdms (Technical Data Management Streaming) files and basic signal processing (peak detection and heart rate calculation). Based on brief experimentation we employed following configuration when conducting measurements described in next section. Measurement with this configuration provided the highest quality signals where noise and movement artefacts were significantly suppressed. Sample rate 600 Sa/s Ground and reference inputs for all groups interconnected to single potential. Bandpass 8th order Butterworth approximation filter with low cutoff frequency fL = 0,5 Hz and high cutoff frequency fH = 60 Hz. Notch 4th order Butterworth approximation filter with fL = 48 Hz and fH = 52 Hz Bipolar derivation between sensing electrodes (Fig. 1 white) and selected reference sensing electrode (Fig. 1 blue).
Jindrich Brablik et al. / IFAC PapersOnLine 52-27 (2019) 421–426
424
3.2 Measurement Deployment To ensure feasibility and reproducibility of the experiments and to cover most of the available electrode placements, as depicted in Fig. 1, we designed the universal measurement deployment, see Fig. 3. The design included total of 16 electrodes that were positioned on the body of the pregnant woman – 14 on the abdomen, one on the right ankle, and one on the left thigh. The switch between the individual configurations was performed by changing the inputs of the system. It is important to note that we omitted some of the electrodes that were redundant (i.e. the recordings included almost identical signals). This lowered the number of electrodes in some of the configurations, such as the one used in Taylor et al., 2003.
Fig. 3. Measurement deployment. Sensing electrodes (white 1 – 11), ground reference (black A, C, F), and common reference electrode (blue D, E, F). Four different configurations of the measurement system were tested: 1.
Jezewski. This configuration was used in the ADFECGDB database (Jezewski et al., 2012). It includes 4 SEs (1, 2, 4, and 7), the GND (C) placed on the left tight and the reference (D) on the bottom of the abdomen.
2.
Behar. This configuration was used in the NIFEADB database (Behar et al., 2019). It includes 4 SEs (1, 3, 4, 5, and 7), the GND (A) placed on the left tight and the reference (D) on the bottom of the abdomen.
3.
Vullings. The configuration introduced in the investigation of Vullings (2010) includes 8 SEs (2, 3, 5, 6, 8, 9, 10, and 11), the GND (B) placed on the left waist, and the reference (F) is placed adjacent to the navel.
4.
Taylor. This configuration of Taylor et al. (2003) includes 12 sensing electrodes (1 – 11); the GND (F) is placed adjacent to the navel and the common reference (E) on the right ankle.
Table 1 summarizes the above mentioned configurations and links them to the design of the general measurement deployment introduced in Fig. 3. Table 1: Configurations tested Configuration Name Jezewski Behar Vullings Taylor
Sensing electrodes 1, 2, 4, 7 1, 3, 4, 5, 7 2, 3, 5, 6, 8 – 11 1 – 11
Ground reference C A B F
Reference D D F E
RESULTS The experiments included measurements on a real subject in 34th week of pregnancy. The total of 14 electrodes were positioned on the abdomen in the way to cover most of the commonly used electrode deployment as shown in Fig. 4.
Fig. 4. Measurement deployment in real subject at 34th week of pregnancy. Figures 5 – 8 show the resulting records of each configuration as listed in the Table 1. In each record, the fetal R peaks are marked. In the cases of the Vullings and Taylor configurations, only 5 signals of the best quality were plotted in figures 4 and 5, respectively. The recordings were shifted by 50 uV in order to enhance the clarity of the readings. Besides the ratio between fetal and maternal component, the shape and orientation of the QRS complexes are among the most important factors for the quality of the maternal ECG suppression. For extraction algorithms, the QRS complexes should be either negative or positive. Contrary, if the QRS complex is biphasic, the algorithms show lower performance. In the most of the signals in the Jezewski configuration, the magnitude of the fetal component is low in comparison with the maternal one (see Fig. 5). However, in the signal recorded by means of SE 4 and 7, the ratio between the fetal and maternal magnitude is low. Additionally, their orientation is the same (positive), and thus, the mECG suppression is feasible. In the Behar configuration (see Fig. 6), the magnitude of the fetal component is comparable with the maternal one, especially in the abdominal signal recorded by means of sensing electrode 4, 5, and 7. Moreover, the placement of the sensing electrodes is convenient for the implementation of the configuration into the clinical practice since it offers a
Jindrich Brablik et al. / IFAC PapersOnLine 52-27 (2019) 421–426
425
space for the placement of the conventional CTG probes if needed. It is also feasible to be incorporated into a wearable device (e.g. using belts or patch system).
Fig. 5. The output signals of the sensing electrodes 1, 2, 4, and 7 (as listed in the Table 1) for the Jezewski configuration.
Fig. 6. The output signals of the sensing electrodes 1 – 5 (as listed in the Table 1) for the Behar configuration.
Fig. 7. The output signals of the sensing electrodes 5, 6, 7, 8, and 4 (as listed in the Table 1) for the Vullings configuration.
Fig. 8. The output signals of the sensing electrodes 2, 3, 5, 9, and 7(as listed in the Table 1) for the Taylor configuration. Table 2: Tested configurations
The quality of the recordings in the Vullings configuration can be considered as the lowest (see Fig. 7). Only 2 out of the 8 SE, i.e. the SE electrodes 5 and 6, are sufficient in terms of the fetal component magnitude. In the others, the fetal component is nearly not notable by the naked eye. The orientation of the maternal QRS complex is positive in the upper SEs (5 – 8), for the rest of them, the orientation is biphasic or mostly negative. The magnitude of the maternal QRS complex in the Taylor configuration is highest among all of the tested configurations (see Fig. 8). The magnitude of the fetal component is significantly lower and the maternal QRS complexes are biphasic. Thus, the fetal ECG extraction would be challenging. The above analysis of each record is summarized by Table 2. The results imply that Behar configuration is the most suitable in terms of the number of SE with a high quality signal, maternal:fetal magnitude ratio (m:f ratio). The only slight negative of this configuration is the morphology of the maternal QRS (mQRS) complex, which is biphasic in some of the signals, which could influence the fetal ECG extraction.
Configuration Name Jezewski Behar Vullings Taylor
Best quality signal SE 3, 4
m:f ratio
mQRS
Low
Positive Negative/ biphasic Positive biphasic
2, 3, 4
Low
5, 6 2, 3, 9, 7
High High
CONCLUSION This article investigated the influence of system configuration on the quality of NI-fECG monitoring. The main factors that were considered and tested were the electrode placement and the hardware system configuration. The results showed that the electrode placement herein denoted as Behar is the most suitable in terms of the signal quality and the clinical feasibility. Moreover, the paper introduced the optimal system configuration to ensure high quality input signals. This was necessary step in order to develop an embedded system for fetal monitoring designed as a wearable device for pregnant subjects. In the future research, it is necessary to test the configurations as well as the performance of the extraction algorithms for the different
426
Jindrich Brablik et al. / IFAC PapersOnLine 52-27 (2019) 421–426
subject, in different stage of pregnancy and different fetal positions. ACKNOWLEDGMENT This article was supported by the Ministry of Education of the Czech Republic (Project No. SP2019/85). This work was supported by the European Regional Development Fund in the Research Centre of Advanced Mechatronic Systems project, project number CZ.02.1.01/0.0/0.0/16 019/0000867 within the Operational Programme Research, Development and Education. This work was supported by the European. Regional Development Fund in A Research Platform focused on Industry 4.0 and Robotics in Ostrava project, CZ.02.1.01/0.0/0.0/17 049/0008425 within the Operational Programme Research, Development and Education. This paper has been elaborated in the framework of the grant programme ”Support for Science and Research in the MoraviaSilesia Region 2018" (RRC/10/2018), financed from the budget of the Moravian-Silesian Region. REFERENCES Behar, J. A., Bonnemains, L., Shulgin, V., Oster, J., Ostras, O., & Lakhno, I. (2019). Noninvasive fetal electrocardiography for the detection of fetal arrhythmias. Prenatal diagnosis, 39(3), 178-187. Behar, J., Andreotti, F., Zaunseder, S., Li, Q., Oster, J., & Clifford, G. D. (2014). An ECG simulator for generating maternal-foetal activity mixtures on abdominal ECG recordings. Physiological measurement, 35(8), 1537. Carrera, J. M. (2003). The technological development of fetal surveillance: a long history. In Controversies in Perinatal Medicine (pp. 22-39). CRC Press. Carrera, J. M. (2003). The technological development of fetal surveillance: a long history. In Controversies in Perinatal Medicine (pp. 22-39). CRC Press. Euliano, T. Y., Nguyen, M. T., Darmanjian, S., McGorray, S. P., Euliano, N., Onkala, A., & Gregg, A. R. (2013). Monitoring uterine activity during labor: a comparison of 3 methods. American journal of obstetrics and gynecology, 208(1), 66-e1. Graatsma, E. M. (2010). Monitoring of fetal heart rate and uterine activity (Doctoral dissertation, Utrecht University). Hayes-Gill, B. R., Barratt, C. W., & Pieri, J. F. (2014). U.S. Patent No. 8,880,140. Washington, DC: U.S. Patent and Trademark Office. Hon, E. H., & Hess, O. W. (1957). Instrumentation of fetal electrocardiography. Science, 125(3247), 553-554. Hon, E. H., & Quilligan, E. J. (1996). The classification of fetal heart rate. Childbirth: The medicalization of obstetrics, 2, 339. Hunter, J. C. A., Braunlin, R. J., Lansford, K. G., & Knoebel, S. B. (1964). U.S. Patent No. 3,120,227. Washington, DC: U.S. Patent and Trademark Office. Huntingford, P. J., & Pendleton, H. J. (1969). The clinical application of cardiotocography. BJOG: An International Journal of Obstetrics & Gynaecology, 76(7), 586-595. Christov, I., Simova, I., & Abacherli, R. (2013, September). Cancellation of the ma-ternal and extraction of the fetal
ECG in noninvasive recordings. In Computing in Cardiology Conference (CinC), 2013 (pp. 153-156). IEEE. Jacod, B. C., Graatsma, E. M., Van Hagen, E., & Visser, G. H. (2010). A validation of electrohysterography for uterine activity monitoring during labour. The Journal of Maternal-Fetal & Neonatal Medicine, 23(1), 17-22. Jezewski, J., Wrobel, J., Matonia, A., Horoba, K., Martinek, R., Kupka, T., & Jezewski, M. (2017). Is abdominal fetal electrocardiography an alternative to doppler ultrasound for FHR variability evaluation?. Frontiers in physiology, 8, 305. Jezewski, J., Matonia, A., Kupka, T., Roj, D., and Czabanski, R. ”Determination of fetal heart rate from abdominal signals: evaluation of beat-to-beat accuracy in relation to the direct fetal electrocardiogram,” Biomedizinische Technik/Biomed. Eng., vol. 57, no. 5, pp. 383- 394, 2012. Kahankova, R., Martinek, R., Jaros, R., Behbehani, K., Matonia, A., Jezewski, M., & Behar, J. A. (2019). A Review of Signal Processing Techniques for NonInvasive Fetal Electrocardiography. IEEE Reviews in Biomedical Engineering. Martinek, R., Kelnar, M., Koudelka, P., Vanus, J., Bilik, P., Janku, P., ... & Zidek, J. (2016). A novel LabVIEWbased multi-channel non-invasive abdominal maternalfetal electrocardiogram signal generator. Physiological measurement, 37(2), 238. Reinhard, J., Hayes-Gill, B. R., Yi, Q., Hatzmann, H., & Schiermeier, S. (2010). Comparison of non-invasive fetal electrocardiogram to Doppler cardiotocogram dur-ing the 1st stage of labor. Journal of perinatal medicine, 38(2), 179-185. Rooijakkers, M. J., Song, S., Rabotti, C., Oei, S. G., Bergmans, J. W., Cantatore, E., & Mischi, M. (2014). Influence of electrode placement on signal quality for ambulatory pregnancy monitoring. Computational and mathematical methods in medicine, 2014. Sartwelle, T. P. (2012). Electronic fetal monitoring: a bridge too far. Journal of Legal Medicine, 33(3), 313-379. Smyth, C. N., & Farrow, J. L. (1958). Present place in obstetrics for foetal phonocardiography and electrocardiography. British Medical Journal, 2(5103), 1005. Smyth, C. N. (1953). Experimental electrocardiography of the foetus. The Lancet, 261(6771), 1124-1126. Taylor, M. J., Smith, M. J., Thomas, M., Green, A. R., Cheng, F., Oseku‐Afful, S., ... & Gardiner, H. M. (2003). Non‐invasive fetal electrocardiography in singleton and multiple pregnancies. BJOG: An International Journal of Obstetrics & Gynaecology, 110(7), 668-678. The MathWorks, Inc. G.tec Support from Data Acquisition Toolbox. MathWorks [online]. 2019. Available at: https://www.mathworks.com/hardware-support/gtec.html Vullings, R. (2010). Non-invasive fetal electrocardiogram: analysis and interpretation. Eindhoven: Technische Universteit Eindhoven.—2010.—225 p. Williams, E. A., & Stallworthy, J. A. (1952). A simple method of internal tocography. The Lancet, 259(6703), 330-332.