EGG Amir

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

10.

2478/v10048-009-0022-6

MEASUREMENT SCIENCE REVIEW, Volume 9, No.5, 2009

Microcontroller -Based System for Electrogastrography


Monitoring Through Wireless Transmission
S. Haddab1, M.Laghrouche1
1
Department of Electronics, Mouloud MAMMERI University, Po Box 17 RP 15000, Tizi Ouzou, Algeria.
shaddab@yahoo.fr, larouche_67@yahoo.fr

Electrogastrography (EGG) is a non-invasive method for recording the electrical activity of the stomach. This paper presents a
system designed for monitoring the EGG physiological variables of a patient outside the hospital environment. The signal acquisition is
achieved by means of an ambulatory system carried by the patient and connected to him through skin electrodes. The acquired signal
is transmitted via the Bluetooth to a mobile phone where the data are stored into the memory and then transferred via the GSM
network to the processing and diagnostic unit in the hospital. EGG is usually contaminated by artefacts and other signals, which are
sometimes difficult to remove. We have used a neural network method for motion artefacts removal and biological signal separation.
Keywords: Electrogastrography, wireless, telemedicine

1. INTRODUCTION

T ELEMEDICINE has been proposed as a means of


improving the efficiency of the healthcare system,
reducing the cost and stress caused by the hospital
This omnipresent slow wave is roughly sinusoidal and
typically identified by its slow frequency and low amplitude
(between 100 and 500 µV). The dominant frequency of EGG
environment1,2. The traditional way of providing telemedicine is 0.05 Hz or 3 cycles per minute (cpm) in healthy humans.
services is to transmit biomedical signals from a patient to a Abnormalities in the regularity of the gastric slow wave have
hospital using “landlines,” such as the Public Switched been frequently reported to be associated with gastric motor
Telephony Network and the Integrated Services Digital disorders and gastrointestinal symptoms, such as nausea and
Network3-5. vomiting, anorexia, dyspepsia and eating disorders.
The development of mobile technology has led to new m- Disturbances of the BER (gastric dysrhythmias) can occur in
Health applications in healthcare provision6. Although face-to- different patterns13 as follows: an increase in the dominant
face consultations between a clinician and a patient will never frequency of the myoelectrical activity of the stomach from 3
be replaced, there are medical cases that can be managed more cpm to 4–9 cpm regular activity is defined as tachygastria; a
efficiently by adopting wireless telemedicine. A wireless decrease in the dominant frequency of the myoelectrical
transmission is becoming increasingly popular in healthcare activity of the stomach from 3 cpm to 1–2 cpm regular activity
and biomedical engineering6. Modern data transmission is defined as bradygastria.
wireless technologies give new possibilities to the designers of EGG is very vulnerable to motion artefacts14 which are not
mobile medical equipment. In the last years a new device only strong, but also have a broadband spectrum, and their
called Bluetooth wireless terminal has appeared on the market. frequencies overlap with that of the gastric myoelectrical
In most cases it is a pair of tiny devices, mostly used for serial activity. This makes it difficult to separate them, and
communication link RS-232 replacement. We have developed quantitative analyses of the EGG data are jeopardized.
a small embedded EGG sensor system prototype using Therefore, an efficient solution for automated detection and
Bluetooth technology. This system is versatile and easily suppression of motion artefacts in the EGG is required. A
transportable. method using feature analysis and back propagation neural
Electrogastrography (EGG) is a method of recording network was also developed to detect and then eliminate
gastric electrical activity from cutaneous electrodes placed on automatically motion artefacts in EGG recordings.
the abdominal surface7. The EGG provides information about
the gastric myoelectric frequency and the amplitude or power
of the EGG signal in the normal or abnormal frequency 2. GENERAL PRESENTATION
ranges. Recording of gastric myoelectrical activity was
Our system is represented by its block diagram in Fig.1. It
described by Alvarez8 in 1921. In the mid 1970s
consists mainly of three modules: The patient’s acquisition
electrogastrography gained renewed interest. Technical
and processing board, the medical storage unit in the mobile
improvements and powerful computers made it possible to
phone and the medical control unit in the PC. The
record and analyze gastric myoelectrical activity, using
microcontroller receives the sensor signals from five inputs
automated spectral analysis. The non-invasive nature of the
through the conditioning circuit and stores them temporarily in
technique makes it an attractive tool in both research and
its EPROM memory. The data are then transferred to the
clinical settings. To date EGG has been used to study a wide
mobile phone via the Bluetooth module and finally sent to the
variety of disorders associated with an altered gastric function
9-12 hospital unit via the GSM network. The other possibility is to
. The main component of gastric myoelectrical activity is
store the data at home in a personal computer.
called gastric slow wave or basic electrical rhythm (BER).

122
MEASUREMENT SCIENCE REVIEW, Volume 9, No.5, 2009

low-frequency components is designed to remove the baseline


(or DC) drift and obtain stable EGG recording. Bradygastria
of below 2 cpm can also be recorded by this equipment.

Fig.1 Block diagram of the data-acquisition system.

3. EGG MEASUREMENT SYSTEM

A. Conditioning circuits
A.1. EGG sensor
The electrogastrograph is constructed to measure the
electrical potential between various points of the body. We use Fig.3 Band pass filter circuit
the Ag-Agcl electrode in a standard EGG acquisition system7.
Depending on how the electrode pairs are connected to the B. Central Processing Unit (CPU)
EGG sensor, different waveforms and amplitudes can be
The Data Acquisition System is composed of two main
obtained. Each pair contains unique information of the
subsystems: the programmable digital data acquisition system
stomach activity.
and transducers with their associated circuits. The requirement
of data acquisition system dictated the use of the
microcontroller. We adopted the 16 f 876 microcontroller. It
is principally composed of 5 channels of 10 bit analog-to-
digital converter (ADC), 3 timers and 2 hardware pulse-width
modulation (PWM) generator modules, an universal
synchronous asynchronous receiver transmitter (USART), an
in circuit serial programming interface (ICSPI), 8K FLASH
memory, and a 20MHz clock. This clock generator is very
suitable for working together with the Bluetooth module to
Fig.2 The EGG amplifier circuit fulfil the wireless communication task.

A.2. Instrumentation Amplifier C. The Bluetooth Transmission Module


The instrumentation amplifier chosen for implementation Bluetooth technology is the wireless protocol used in the
is the INA114 from Burr-Brown (Fig.2). The signal amplitude transmitter. Bluetooth has many advantages. First, it is
at the output of the sensor is about 500µV; in this case the available in a user-friendly modular form. There are many
value of Rg is chosen to be 51, making the signal’s gain close available Bluetooth devices which hide the Bluetooth stack
to one thousand. and allow the user to interact with the device using simple
modem commands. This reduces development time
A.3. Filtering considerably. In addition, Bluetooth is a common technology
To remove the unwanted frequencies in this case requires a on mobile phones, and this will expand the range of use for the
highly selective bandpass four order filter. The circuit used is transmitter. The Ezurio Bluetooth Intelligent Serial Module II
made up with four similar stages based on OP 07 operational (BISM II) is the device selected for the wireless module. This
amplifiers mounted in cascade (Fig.3). The signal frequency chip is in compliance with Bluetooth and provides USB and
components at 1.8 cpm and at 16.0 cpm have an attenuation of UART communication interfaces. As shown in Fig.4, this
3 dB or 71% in amplitude. Frequency components of interest system can use UART interface to carry out the data
in the EGG include normal 3.0 cpm slow waves, bradygastria transmission between the module and the microcontroller.
of 0.5 to 2.4 cpm, and tachygastria of 3.7 to 9.0 cpm. From the Any device connected via Bluetooth to the PAP-BT board
frequency response of this filter one can see that normal can thus receive data in real-time. Having an excellent range
gastric slow waves and abnormal tachygastrias will be of up to 300 meters with very low power consumption of less
recorded with the maximum amplification. The attenuation of than 36 mA makes this module a true class leader.

123
MEASUREMENT SCIENCE REVIEW, Volume 9, No.5, 2009

Fig.4 Central processing unit with Bluetooth transmission


Fig.5-b Spectral representation of the network input signal.

4. EXPERIMENTATION AND EGG ANALYSIS In the part of our signal (Fig.5-a) we observe two pulse
An acquisition has been carried out with the ambulatory peaks around 40 seconds and 180 seconds. The suppression of
system on a volunteer patient over a 4 hours time period. The the artefact peaks by filtering is to be avoided because of the
sampling period chosen was 1 second. Thus, the number of presence of a frequency band common to the useful EGG
samples acquired over 4 hours was 14.400. These data values signal and we have chosen to use neural nets to overcome this
were transmitted via the Bluetooth to the cell telephone where difficulty.
they were stored on a Secure Digital memory card using the
FAT 16 format. After 24 hours the stored data were sent to the A . Motion artefact in EGG measurement
PC in the hospital via the web service. Up to a recent past, the deletion of EGG data with motion
Fig.5-a represents a slice of 250 samples of the received artefacts was performed by visual inspection. Unlike other
signal stored in the PC memory. electrophysiological recordings, the EGG is very vulnerable to
This interval has been chosen with artefacts peaks resulting motion artefacts in the ambulatory carried system. The
from the volunteer’s movement during the acquisition. These artefacts are not only strong, but also have a broadband
pulses with very short durations are therefore present inside spectrum, and their frequencies overlap with that of the gastric
the signal received by the PC. The spectral representation of myoelectrical activity. This makes it difficult to separate them,
the signal is given in Fig.5-b. and quantitative analyses of the EGG data are jeopardized.
Therefore, an efficient solution for automated detection and
elimination of motion artefacts in the EGG is required.

B. Suppression of the artefacts by using the neural network


methods
Neural networks have been widely used in pattern
recognition due to their great potential of high performance,
flexibility, cost-effective functionality, and capability for real-
time applications. A number of successful applications of the
neural network to biomedical signal detection and pattern
recognition have been reported 14-15.
In this study, a method using feature analysis and
backpropagation neural network is developed to detect and
then eliminate automatically motion artefacts in EGG
recordings. Special studies on motion artefacts of EGG are
designed. The feature of the EGG is analyzed and used as the
input of the detection system, which is realized by a
backpropagation neural network. Performance of the neural
networks using different features is compared to obtain an
optimal selection of the features. The back propagation neural
Fig.5-a Representation of 250 samples of the signal received network is so powerful that it can learn a mapping of many
by the PC
complexities, and is used in perhaps 80 to 90% of practical
applications. Fig.6 shows the network structure. The network

124
MEASUREMENT SCIENCE REVIEW, Volume 9, No.5, 2009

has three layers: one input layer, one hidden layer, and one
output layer. The output for the data with motion artefacts is
set to one, whereas the output for those without motion
artefacts is set to 0.

Fig.7-b Spectral representation of the network output signal

Fig.6 The network structure 5. DISCUSSION


The previous result shows the quasi-suppression of the
The number of neurons in the hidden layer is one of the artefact peaks relatively to the shape of the acquired EGG
parameters to be designed, and experimental results show that signal. In the time-domain, Fig.5-a and 7-a above represent the
three were optimal. The transfer function of the hidden layer is input and output signals of the net, respectively. We have
a tan-sigmoid function, whereas that of the output layer is a clearly verified the absence of the two artefact peaks that were
hard-limit function8. Because all transfer functions used in present in the input signal of the net.
backpropagation network must be differentiable (derivatives This result is confirmed in the spectral representation of the
of error are calculated at each learning step), a log-sigmoid signal. Indeed, we observe, through Fig.5-b and 7-b a decrease
function is used in the output layer. of the frequency magnitude close to that of the fundamental
gastric slow wave. These frequencies correspond to the
C. Application to the EGG signal artefact peaks that have been suppressed. The BER gastric
After fixing and testing the parameters of the net with wave of a frequency close to 0.05Hz appears clearly in the
different simulations, we applied it to the case of an acquired output signal of the net and the frequency components of the
EGG signal. For this purpose, we selected a pack of 250 movement’s artefacts are highly reduced.
samples that contained two artefact peaks. The obtained This method overcomes the problem of non-adaptation of the
results are presented in the following diagrams (Fig.7-a and classical filtering techniques in the case of this type of noise.
Fig.7-b).
6. CONCLUSION
This study proposes and investigates the application of a
GSM and the Bluetooth technology to a wireless-type EGG
signal measuring system and the feasibility of the proposed
system. According to our experimental results, the proposed
system is proved to be feasible. The non-invasive nature of the
technique makes it an attractive tool, in both the research and
clinical settings. The use of an autonomous ambulatory system
enables the patient to remain free in his movements during the
signal acquisition. Our wireless transmission system permits
to store the data acquired in the patient’s mobile phone
memory, and then to transmit them to the processing unit of
the hospital through the GSM net. The presence of the patient
in the hospital becomes no longer necessary and the distance
separating the patient and the hospital is no longer a
constraint.
Thanks to the ambulatory feature of the acquisition, the
movement artefacts are unavoidable and disturb interpreting
and analysis of the EGG signal. The application of an
Fig. 7-a Output signal of the neural network. automatic processing method based on the use of neural
networks enables a very important reduction of these artefact
peaks.

125
MEASUREMENT SCIENCE REVIEW, Volume 9, No.5, 2009

REFERENCES [8] [Alvarez, W.C. (1922). The electrogastrogram and what


[1] Tura, A., Badanai, M., Longo, O., Quareni, L. (2003). A it shows. Journal of the American Medical Association,
medical wearable device with wireless Bluetooth based 78, 1116-19.
data transmission. Measurement Science Review, 3, 1-4. [9] Lin, Z., Eaker, E.Y., Sarosiek, I., McCallum, R.W.
[2] Rasid, M.F.A., Woodward, B. (2005). Bluetooth (1999). Gastric myoelectrical activity and gastric
telemedicine processor for multichannel biomedical emptying in patients with functional dyspepsia.
signal transmission via mobile cellular networks. IEEE American Journal of Gastroenterology, 94, 2384–2389.
Transactions on Information Technology in [10] Verhagen, M.A.M.T., van Schelven, L.J., Samsom, M.,
Biomedicine, 9 (1), 35-43. Smout, A.J.P.M. (1999). Pitfalls in the analysis of
[3] Chien, J.-R.C., Tai, C.C. (2004). A wireless Bluetooth electrogastrographic recordings. Gastroenterology, 117,
device applied to a non-contact type breathing 453–460.
monitoring system. In IEEE International Conference on [11] Huizinga, J.D., Robinson, T.L., Thomsen, L. (2000). The
Networking, Sensing and Control, March 21-23, 2004, search for the origin of rhythmicity in intestinal
Vol. 1, 172-173. contraction; from tissue to single cells.
[4] Chien, J.-R.C., Tai, C.C. (2006). Handheld Neurogastroenterology and Motility, 12 (1), 3-9.
electrocardiogram measurement instrument using a new [12] Hubka, P., Rosík, V., Ždiňák, J., Tyšler, M., Hulín, I.
peak quantification method algorithm built on a system- (2005). Independent component analysis of
on-chip embedded system. Review of Scientific electrogastrographic signals. Measurement Science
Instruments, 77 (9), 095106. Review, 5, 21-24.
[5] Woodward, B., Istepanian, R.S.H., Richards, C.I. (2001). [13] Simonian, H.P., Panganamamula, K., Chen, J.Z., Fisher,
Design of a telemedicine system using a mobile R.S., Parkman, H.P. (2003). Multichannel electro -
telephone. IEEE Transactions on Information gastrography (EGG) in symptomatic patients: A single
Technology in Biomedicine, 5 (1), 13–15. center study. Neurogastroenterology and Motility, 15,
[6] Rubel, P., Fayn, J., Nollo, G., Assanelli, D., Li, B., 338.
Restier, L., et al. (2005). Toward personal health in [14] Liang, J., Cheung, J., Chen, Z. (1997). Detection and
cardiology. Results from the EPI-MEDICS telemedicine deletion of motion artefacts in electrogastrogram using
project. Journal of Electrocardiology, 38 (4), 100-106. feature analysis and neural networks. Annals of
[7] Haddab, S., Bouchoucha, M., Cugnenc, P.-H., Barbier, Biomedical Engineering, 25, 850-857.
J.-P. (1990). New method for electro gastrographic [15] Chen, J., Lin, Z., Wu, Q., McCallum, R.W. (1995). Non-
analysis. In Proceedings of Third Annual IEEE invasive identification of gastric contractions from
Symposium on Computer-Based Medical Systems, June surface electrogastrogram using back-propagation
3-6, 1990, 418-425. neural networks. Medica Engineering and Physics, 17,
219-225.

126

You might also like