Adaptive Processing of Bioelectric Abdominal Signals To Improve

the Reliability of Fetal Home Telemonitoring

TOMASZ KUPKA, JANUSZ JEZEWSKI, ADAM MATONIA, DAWID ROJ, KRZYSZTOF HOROBA

Department of Biomedical Signal Processing

Institute of Medical Technology and Equipment

Roosevelta 118, 41-800 Zabrze,

POLAND

tomekk@itam.zabrze.pl

Abstract: - Centralized system for fetal home telemonitoring with wireless data transmission based on GSM service

and internet is presented in this paper. Cardiotocographic signals are provided by mobile instrumentation which

consists of bioelectrical signal recorder and Tablet PC. It enables to support a classical interpretation of

cardiotocographic trace by information on fetal electrocardiogram morphology. The system enables analysis, dynamic

presentation and archiving of acquired signals and medical data. Novelty of proposed approach relies on modification

of the procedures to process the abdominal signals in mobile instrumentation and adaptive controlling of the

monitoring session in surveillance center. These adaptations are performed automatically through advanced algorithms

based on continuous analyzing of the quality and quantitative parameters of the acquired signals. In that way the

amount and content of data transmitted through remote channels to surveillance center can be controlled to ensure the

most reliable assessment of fetal well-being.

Key-Words: - home monitoring, telemedicine, high-risk pregnancy, fetal surveillance, FECG, signal processing

1 Introduction Cardiotocography (CTG) is commonly used method of

fetal monitoring, which enables evaluation of the fetal

wellbeing during pregnancy and in labour. The method

relies on the analysis of characteristic fetal heart rate

(FHR) patterns in relation to the uterine contractions

(UC) and fetal movements. The normal heart activity

indicates the adequacy of fetal oxygenation and correct

functioning of central nervous system. Usually the FHR

signal acquisition is based on the Doppler ultrasound

technique which records the mechanical activity of the

fetal heart. Determination of instantaneous FHR relies

on the detection of heart beats based on the analysis of

ultrasound beam reflected from the moving valves or

walls. The main advantage of the Doppler ultrasound

technique is simplicity of application and non-

invasiveness, although the accuracy is rather low for the

automated signal analysis comprising evaluation of FHR

variability at a level of single heart beats. The

mechanical method as an indirect measuring technique

records the effects of electric excitation i.e. fetal heart

movement. Considerably higher accuracy and reliability

can be obtained using the primary bioelectric signal –

fetal electrocardiogram (FECG). Consecutive cardiac

cycles can be determined more accurately through

detection of QRS complexes in fetal electrocardiogram

then by analyzing the reflected ultrasound beam of a

complex shape. Authors developed measurement

instrumentation for acquisition and analysis of the fetal

electrocardiogram and uterine contraction activity on a

basis of bioelectrical signals recorded from maternal

abdominal wall. This Mobile Instrumentation (MI)

provides the telemedical system being under

development with both the FHR and UC signals.

The signal processing procedures running in Mobile

Instrumentation are preliminary adapted to measurement

conditions being changed. The proposed strategy is

based on the estimated quality of the biosignals recorded

from abdominal wall of the pregnant women.

Additionally, in the Surveillance Center a detailed

analysis of the FHR signal loss is carried out to control

an adaptive modification of the biosignal interpretation

algorithms in the MI. The fetal state is assessed by

means so called non-stress test (NST), whose result

decides about further diagnostic procedures.

In medical centers, a need for simultaneous

monitoring of many patients leads to wide use of

centralized fetal surveillance systems [1]. Recorded

signals from all fetal monitors along with analysis results

are simultaneously presented on the monitoring station

in a form of graphical and numerical data. Database

contains the archive of traces, analysis results and

medical history of patients. The limitation of currently

used systems is a lack of possibility to monitor the

patient outside the hospital. So far continuous medical

care requires a hospitalization of pregnant woman even

if there is no direct risk for patient’s health. It results in

high cost of longer hospital stay and discomfort for a

patient. The optimal solution seems to be a remote fetal

LATEST TRENDS on COMPUTERS (Volume I)

ISSN: 1792-4251 196 ISBN: 978-960-474-201-1

monitoring at patient’s home. Patients with high-risk

pregnancy as well as with post-term pregnancy are

particularly predisposed to cyclic home monitoring

sessions for follow-up of the fetal development process

[2].

When monitoring is based on recording of

bioelectrical abdominal signals, very careful preparation

of patient’s skin is required to obtain the signal of good

quality. Additionally, in the first stage of monitoring,

when template maternal QRS complexes are created,

patient should not move. This is the reason that the home

monitoring procedure should be carried out by the

hospital patient’s care staff. The operator with a mobile

fetal monitoring instrumentation visits particular patients

appointed to be monitored according to a fixed schedule.

However, some logistic problems with visit scheduling,

especially for large medical centers incorporating

numerous patients, should be solved.

2 System structure Fetal centralized surveillance system can be defined as a

set of fetal monitoring bedside devices, which transmit

cardiotocographic data to the central computer. In case

of telemonitoring the remote channel is assumed to work

on-line, so the monitoring session will be carried out in

the real-time. If a continuous data transmission is

assured, the monitoring can run in the same way as if the

patient would be in hospital (Fig.1). Wireless

communication is based on the data transmission service

used in cellular phones system, and internet which is

used for data transfer between GSM network and the

Surveillance Centre [3]. SC includes TCP/IP interface

allowing communicating with Mobile Instrumentation

via Internet. The system software has to be extended to

assure continuous, error-free data transmission and to

enable communication with the patient.

Mobile Instrumentation comprises a bioelectrical

signal recorder and Tablet Personal Computer or the

Personal Digital Assistant (PDA) as a computer with

built–in GSM module assuring the wireless connection

through the internet [4]. The computer software enables

acquisition and processing the signals from recorder,

their dynamic presentation and on-line evaluation of

their quality. Additionally, all the data acquired from the

fetal recorder are converted into the format established in

the surveillance system.

Fig. 1 The structure and interactions between different components in telemedical system

for home fetal monitoring with online adaptive analysis of bioelectrical abdominal signals.

LATEST TRENDS on COMPUTERS (Volume I)

ISSN: 1792-4251 197 ISBN: 978-960-474-201-1

The recorder is equipped with four differential

channels for measurement of abdominal signals. Typical

configuration comprises four electrodes placed around

the navel and the reference electrode placed above the

pubic symphysis. Additionally, the common mode

reference electrode is placed on the left leg. The

necessity of using four abdominal leads results from the

fact that very often the FECG signal of good quality is

present only in one lead. Considering abdominal FECG,

the basic merit of the presented recording unit is a very

low level of its own noise which does not exceed 1 µV

(peak-to-peak) measured with reference to the inputs,

and a high value of CMRR coefficient (120 dB). These

parameters have been obtained thanks to novel recorder

electronic circuits structure including complete

separation of analog part from the digital one [5]. Each

channel is equipped with an amplifier with gain control

that allows the amplification of recorded signals from

the tens of microvolts up to the level of several volts.

The band-pass filter removes low frequency components

and thus prevents the reaching of saturation state by

amplifiers in case of strong isoline drift.

The surveillance system located in hospital has a

capability of simultaneous monitoring of up to 24

patients, both remotely and within hospital. The main

system tasks are: the analysis of incoming data, dynamic

presentation of traces along with analysis results as well

as storing and printing the data. The quantitative

parameters describing acquired signals are used to detect

alerting situations.

Fig. 2 Computer screen of a surveillance centre presenting an

enlarged cardiotocographic trace from one of the patients

(scale 1 cm/min). The upper signal is the fetal heart rate, the

lower one is the uterine activity and markers between them

represent detected fetal movements. The horizontal bars

directly above the waveforms identify the characteristic trace

patterns. On the right side there is a window presenting last

three averaged fetal P-QRS-T complexes and corresponding

values of T/QRS coefficient.

The form of information displayed by the system

should not affect the interpretation of CTG trace. Since

in classic cardiotocography the acquired signals are

visualized as waveforms printed on thermosensitive

paper, the display provides the same graphic forms with

regard to quality, aspect ratio and waveforms flowing

(Fig.2). For records which are provided by bioelectrical

recorder an additional window can be displayed,

containing last three averaged fetal P-QRS-T complexes

and corresponding values of T/QRS coefficient. In

addition, any time-amplitude relationships can be

measured and stored in the database with appropriate

comment.

Optional workstation, connected through the local

network, provides an instant access to patients data and

acquired signals. Workstation can be used to set up the

system, to create paper documentation as well as to

process the signals recorded in the off-line mode (e.g. in

case of total breaking of the communication link). It is

possible to access the information stored in the archive

from outside the hospital via internet. This feature allows

the obstetrician to view the monitoring records at any

time he needs. However, due to the personal data

protection, the access is permitted only for attending

doctor for a given patient.

2.1 Signal analysis in Mobile Instrumentation The signal analysis is aimed at determination of fetal and

maternal heart rate signals (FHR and MHR), uterine

contractile activity (UC) from electrical activity of

uterine muscle - electrohysterogram (EHG), as well as

averaged P-QRS-T complex from fetal

electrocardiogram (Fig. 3).

Signal recorded from maternal abdomen includes the

maternal (MECG) and the fetal (FECG)

electrocardiograms, EHG as well as many unwanted

muscle and low frequency components. Suppression of

the dominating component in the abdominal signal –

maternal electrocardiogram – is the first, and at the same

time, the decisive step in abdominal fetal

electrocardiography [2, 6]. At first, the spatial filtering

based on the generalized singular value decomposition

(GSVD) is applied to extract pure dominating maternal

electrocardiogram from abdominal signal [7]. Having

such MECG the maternal QRS complexes can be

detected very precisely. Then, information on maternal

QRS complexes localization is used to determine the

maternal heart rate and to suppress the maternal ECG in

the abdominal signal which makes possible further

detection of the fetal QRS complexes. The basic

approach to MECG suppression is blanking, where

suitably long segment of the abdominal signal

comprising maternal complex is simply replaced by

isoline values. Unfortunately, in case of coincidence of

LATEST TRENDS on COMPUTERS (Volume I)

ISSN: 1792-4251 198 ISBN: 978-960-474-201-1

maternal and fetal complexes the latter one is rejected

causing partial FHR signal loss [6]. Nevertheless, this

disadvantage does not affect clinician’s interpretation

since for classical visual analysis the FHR signal is

averaged over 2.5-second periods. Additionally, low

computational complexity of the blanking approach

leads to significant reduction of power consumption in

Tablet or PDA computers.

Fig. 3 Adaptation of the biosignal analysis algorithms in the

MI controlled by the FECG quality index (SQI).

Blanking is applied to the abdominal signal in every

acquisition channel, with simultaneous controlling of the

quality of final FECG signal. The signal quality index

(SQI) takes values from 0 to 3, where 0 means very

weak signal, whereas 3 – its best quality. If the SQI

reaches the value of three in particular channel then the

fetal QRS complexes detection and consecutively

calculation of Trr intervals are carried out using the

signal acquired through this channel. The detection

function relies on matching filtering and application of a

set of decision rules [8]. Only these FHR values which

fulfill physiological criteria are finally accepted as

correct ones. If the SQI takes a value of two (satisfying

signal quality) in the best channel, then additional noise

suppression based on projective filtering [9] is applied

before QRS detection starts. If none of the channels

provides satisfying quality of the signal (2 or 3) this

means that the one-channel detection with blanking does

not ensure good results which causes significant FHR

loss. In that case more precise and advanced maternal

ECG suppression method has to be used, i.e. the method

based on subtraction of appropriately rescaled and

adaptively modified the reference maternal P-QRS-T

complex [2]. Suppression takes place in every abdominal

channel, and thus the fetal QRS detection is

multichannel. In this approach the additional noise

removal procedure is applied to improve FECG quality

before detection process starts. Any channel with SQI

equal to 0 is excluded from the fetal QRS detection.

Surveillance Center decides about a way of analysis

of the recorded signals to evaluate the fetal state.

Primary, an interpretation of CTG records is carried out.

If its result is unclear the additional morphology analysis

of fetal P-QRS-T complex is performed. Then, the

advanced suppression algorithm basing on subtraction of

reference maternal P-QRS-T complex is involved

automatically. Consecutive fetal P-QRS-T complexes

obtained in such way undergo the weighted averaging

[5]. For consecutive averaged complexes the relation of

the amplitude of T wave to the amplitude of QRS

complex is calculated (T/QRS ratio). Averaged

complexes together with relating T/QRS values are sent

to the Surveillance Center, where their further analysis is

carried out.

The contractile activity signal is determined basing

on electrical uterine muscle activity. For this task, the

abdominal signals are fed to low-pass filter with cut-off

frequency of 3.5 Hz, which corresponds to

electrohysterogram frequency band. The signals are then

downsampled from 500 Hz to 10 Hz. In the next stage

the resulting signal of the contractile activity is obtained

by composition of the signals from four leads. After that

the RMS values are calculated in the window of 60-s

width shifted with 3 seconds to obtain the consecutive

values of the UC signal [10].

2.2 Signal analysis in Surveillance Center In the Surveillance Center the algorithms have been

implemented to analyze CTG signal being received. The

analysis performed is consistent with guidelines of the

FIGO Subcommittee on Standards in Perinatal Medicine

[11]. The CTG analysis is accomplished in a number of

stages. In the first stage the FHR signal is verified basing

on identification and elimination of artifacts. Then the

analysis of signal loss and averaging over 2.5-s period

with interpolation of the lost values are carried out. It is

LATEST TRENDS on COMPUTERS (Volume I)

ISSN: 1792-4251 199 ISBN: 978-960-474-201-1

crucial for estimation of the FHR baseline – a basis for

recognition of the acceleration and deceleration patterns

as well as tachycardia and bradycardia episodes. In

addition, a set of indices is determined to estimate an

long and short term instantaneous FHR variability [5].

These indices are calculated for each one-minute

segment basing on a heart beat marker obtained from

FECG. Analysis of the uterine contraction activity signal

is aimed at recognition of the contraction episodes in

relation to so called basal tone, and then determination

of parameters describing contractions. They are very

important for recognition and classification of the

deceleration patterns, whereas information on fetal

movement activity is crucial for acceleration patterns.

Fig. 4 Adaptation of the biosignal analysis algorithms in the

Surveillance Center.

The FHR signal loss evaluated in Surveillance Centre

is key element for adaptive control of the algorithms for

analysis of the biosignals recorded in Mobile

Instrumentation (Fig. 4). But, preliminary adaptation

takes place automatically in Mobile Instrumentation on a

basis of the FECG quality index. Depending on its value,

less or more advanced method is selected to process the

bioelectrical signals recorded from maternal abdomen. In

the Surveillance Center the loss level of the FHR signal

(evenly sampled at 4 Hz and averaged over 2.5-second

segments) is continuously evaluated and checked.

When FHR loss exceeds 10% for at least two

minutes, the message is sent to MI which informs the

operator to check contact between electrodes and skin,

current position of patient (left side is recommended

during CTG monitoring session) as well as measurement

conditions. If the FHR loss signal remains above 10%

for the next five minutes the message is sent

automatically to Mobile Instrumentation in order to

execute more advanced processing of recorded

biosignals. If despite this action the FHR loss level is not

lower than 10% for next five minutes then alerting

message is displayed in Surveillance Center. It should be

taken into account by clinician during assessment of

CTG record since significant FHR signal loss may affect

his interpretation. As long as the FHR loss is below 10%

the selection of appropriate signal processing algorithms

is controlled in Mobile Instrumentation by fetal ECG

quality index evaluated.

Apart from procedures to control algorithms of

biosignal processing, an adaptive controlling of the CTG

monitoring session has been implemented in SC (Fig. 5),

according to non-stress test. This name emphasizes the

fact that NST is fully noninvasive. When computer-

aided CTG monitoring system is used this test is based

on detailed analysis of CTG record and provides

qualitative evaluation of fetal state. Procedure of

automated non-stress test leads to determination of test

result (reactive, nonreactive or suspicious) by means of a

set of decision rules applied to parameters of quantitative

CTG analysis. Usually, the test is carried out for at least

30 minutes.

If NST is reactive the CTG monitoring is finished,

because fetal well-being has been confirmed. In case of

any other test result, the test can be prolonged by next 30

minutes, however it is a decision made by doctor on call.

This is justified because the fetal activity can be

significantly lower during fetal sleep phase.

Fig. 5 Adaptive controlling the CTG monitoring in the

Surveillance Center.

LATEST TRENDS on COMPUTERS (Volume I)

ISSN: 1792-4251 200 ISBN: 978-960-474-201-1

The other approach relies on providing averaged fetal

P-QRS-T complex and T/QRS ratio that are sent from

Mobile Instrumentation on a request from Surveillance

Center. The aim of this additional information is to help

a clinician in making appropriate decision concerning

further patient’s treatment and necessity of her

hospitalization.

3 Conclucions The proposed system of fetal monitoring will certainly

improve patient’s comfort and reduce the cost of medical

care. Additionally, instant access to database will make

the communication between the patient and her attending

doctor much easier.

Automated and adaptive adjusting of the algorithms

applied to process recorded biosignals as well as

adaptive controlling the CTG monitoring session enable

considerable reduction of amount of data that have to be

sent to Surveillance Center from particular Mobile

Instrumentation. In most cases, the simplest algorithms

and regular duration (30 minutes) of CTG monitoring

session are sufficient. It is very important considering

many tasks that have to be accomplished by Surveillance

Centre simultaneously managing many patients. In

Mobile Instrumentation the proposed approach leads to

lengthening of the battery operating time and reducing

the cost of transmission through GSM/Internet.

Application of the bioelectrical signals recorder

enables information on fetal heart rhythm to be obtained

in a form of event series (heart beats). This enables more

precise calculation of the FHR instantaneous variability

indices, which is carried out in the Surveillance System.

Additionally, very important part of information is

provided by analysis of the averaged fetal P-QRS-T

complex, particularly with evaluation of the T/QRS

value changes. This information enables verification of

suspicious cardiotocographic traces.

Acknowledgement:

Scientific work financed from the State Committee for

Scientific Research resources in 2007-2011 years as a

research project No. N518 335935.

References:

[1] Jezewski J, Wrobel J, Horoba K, et al, Centralised

fetal monitoring system with hardware-based data

flow control, Proc III Conf MEDSIP, 2006,

pp. 51-54.

[2] Jezewski J, Matonia A, Kupka T, Gacek A, Horoba

K, Suppression of maternal ECG interference in

abdominal fetal electrocardiogram, IFMBE Proc of

the 12th Nordic Baltic Conference on Biomedical

Engineering and Medical Physics, Iceland, 2002,

pp.162-163.

[3] Salvador CH, Carrasco MP, et al. Airmed-Cardio, A

GSM and Internet Services-Based System for Out-of-

Hospital Follow-Up of Cardiac Patients, IEEE Trans

Inf Technol Biomed, 2005, Vol. 9, pp. 73-85.

[4] Hod M, Kerner R, Telemedicine for antenatal

surveillance of high-risk pregnancies with

ambulatory and home fetal rate monitoring: an

update, J Perinat Med, Vol. 31, 2003, pp. 195-200.

[5] Gacek A, Matonia A, Jezewski J, Kupka K,

Recognition of early symptoms of fetal distress with

on-line analysis of bioelectrical signals from

maternal abdomen, Biocybernetics and Biomedical

Engineering, 2007, Vol. 27(1), pp. 207-216.

[6] Kupka T, Jezewski J, Matonia A, Wrobel J, Horoba

K, Coincidence of maternal and fetal QRS complexes

in view of fetal heart rate determination, J Med

Inform Technology, Vol. 4, 2002, pp. 49-55.

[7] Callaerts D, DE Moor B, Vandewalle J, Sansen W,

Comparison of SVD method to extract the fetal

electrocardiogram from cutaneous electrode signals,

Med Biolog Eng Comput, 1990, Vol. 28,

pp. 217-224.

[8] Gibson NM, Woolfson MS, Crowe JA, Detection of

fetal electrocardiogram signals using matched filters

witch adaptive normalization, Med Biol Eng Comput,

1997, Vol. 35, pp. 216-222.

[9] Kotas M, Jezewski J, Matonia A, Kupka T, Spatio –

temporal filtering for fetal QRS enhancement,

IFMBE Proceedings of the World Congress on

Medical Physics and Biomedical Engineering,

Munich, IX 2009, Vol. 25, pp. 389-392.

[10] Radhakrishnan N, Wilson JD, Lowery C, et al., A

fast algorithm for detecting contractions in uterine

electromyography, IEEE Eng Med Biol, 2000,

pp. 89-94.

[11] Rooth G, Guidelines for the use of fetal monitoring,

Int J Gynaecol Obstet, 1987, 25, pp. 159-167.

LATEST TRENDS on COMPUTERS (Volume I)

ISSN: 1792-4251 201 ISBN: 978-960-474-201-1