You are on page 1of 5

ECG Sonification: A New Approach for Diagnosis of Cardiac

Pathologies
Shaid Hasan Iqbal Kabir Pratic A Muntakim
Electrical and Electronic Electrical and Electronic Electrical and Electronic
Engineering Engineering Engineering
Bangladesh University of Bangladesh University of Bangladesh University of
Engineering and Technology Engineering and Technology Engineering and Technology
Dhaka Bangladesh Dhaka Bangladesh Dhaka Bangladesh
hasan.s.eeebuet@gmail.com iqbalkabir5004@gmail.com praticbueteee@gmail.com

ABSTRACT
Easy detection of electrocardiogram (ECG) data is highly required 1 Introduction
in modern clinical system in case of different diseases. The Human listening system is very powerful that can easily identify
present existing technique to represent and analysis data is sources of sound, spoken words even under noisy situation. Even
visualization. Another alternative way of data representation in the modern technology, computer can’t reproduce this
known as sonification can make a revolutionary development in inimitable recognition task that our listening system do. Besides,
many clinical applications. In this work, we have applied human is capable of interpreting sounds to extract actual meaning
sonification technique on ECG dataset and demonstrated a user of the words. Our brain can produce useful characterized data if
study on 20 undergraduate students for diagnosis of cardiac enough opportunities are provided to train and hear the data [1].
pathologies. We have also made a user study comparison between Gregory Kramer founded the International Community for
sonification and visualization technique. Our study can be the Auditory Display (ICAD) in 1992 and then data sonification was
foundation in further sonification and medical researches. firstly introduced [1]. The term sonification means the
transformation of data into sound. In 1999, data transformation
CCS CONCEPTS into acoustic signal was reported by Kramer et al. [2]. In the last
• Applied computing~Life and medical sciences • Applied few years, many more researches have been performed on
computing~Health care information systems sonification technique because of easy detection of abnormalities
of the ECG signals [3].
KEYWORDS Previous study shows that, respiratory sinus arrhythmia and
congestive heart failure can easily be identified by hearing the
Sonification, ECG, Cardiac Pathology. sound of the increasing heart rate and loss of the variability of
heart rate, respectively compared to the healthy man’s heart rate
ACM Reference format: sounds [4]. The recent two methods e.g. water ambience
soundscapes and timbre morphing based on sonification are
Shaid Hasan , Iqbal Kabir and Pratic A Muntakim.
introduced by Andrea Lorena Aldana Blanco et. al. in 2016 [5].
ECG Sonification: A New Approach for Diagnosis of Cardiac
Pathologies. In Proceedings of 6th International Conference on
Median bias removal method, peak picking and grouping method,
Networking, Systems and Security (NSysS 2019). December 17–19, 2019,
and linear filtering and non-linear scaling method are applied to
Dhaka, Bangladesh. https://doi.org/10.1145/3362966.3362968 peak the R-wave accurately from the QRS complex [6]. On the
other hand, voice and matlab software are used for image data
sonification which makes a revolutionary change in medicine [7].
Besides, EEG sonification has also been possible to hear sound
musically though it not sounds by nature which is the conversion
of the activities of the brain signal into sound. During
reproduction, it is possible to hear clear sound of EEG signal by
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or
speeding up 20 times [8].
distributed for profit or commercial advantage and that copies bear this notice and
the full citation on the first page. Copyrights for components of this work owned
Today the only existing procedure to analyse ECG data is
by others than ACM must be honored. Abstracting with credit is permitted. To visualization and digital processing which is sometimes unable to
copy otherwise, or republish, to post on servers or to redistribute to lists, requires detect correct ECG pattern in many of the cases if the patient is
prior specific permission and/or a fee. Request permissions from attacked by diseases with the specific one. Sound signals of
Permissions@acm.org.
6th NSysS 2019, December 17–19, 2019, Dhaka, Bangladesh different clinical pathologies carry useful information which helps
© 2019 Association for Computing Machinery us to analyse complexity of the data with scientific computation
ACM ISBN 978-1-4503-7699-0/19/12…$15.00 [9]. A recent work shows that the accuracies of the sonification of
https://doi.org/10.1145/3362966.3362968
the ECG signals are 78%, 68% and 50% for the people having
medical courses with cardiology training or not completed
medical courses without training on cardiology or not having any
medical background [10].
In this work, we used sonification technique to convert ECG data
into sounds to detect cardiac pathologies. We have also
demonstrated a case study on 20 undergraduate students for
diagnosis of cardiac pathologies using sonification technique and
also with ECG graph visualization. A comparison is made
between this two techniques (sonification and visualization) for
diagnosis of cardiac pathologies.

2 Methods
2.1 Dataset
We have selected four cardiac pathologies for our work: Atrial (b)
Fibrillation, Myocardial Infarction, Myocarditis, Coronary Artery.
For pathologies we have used data from “St.-Petersburg Institute
of Cardio logical Technics 12-lead Arrhythmia Database”
(incartdb) on www.physionet.org. For healthy sample “PTB
database” on www.physionet.org was used. The data was
downloaded as .mat format for using in MATLAB software.
From the 12-channel dataset we extracted the first six leads (I,
II, III, aVR, aVL, avF) as shown in Fig 1. For each category (four
pathologies and healthy man) we have used two samples for user
study.

(c)

(a)

(d)
In user study, we have used total ten samples, two samples for
each five categories (four cardiac pathologies and one healthy
man). One sample from each five categories are placed in the
‘Select Sample’ pop up menu and arbitrarily named as Sample 1,
Sample 2, Sample 3, Sample 4 and Sample 5. Other five samples
(ideal sample) are placed in the rectangular shaped buttons in right
side of the GUI. In the user study we mainly investigated whether
user could identify random samples by listening audio sound and
visualizing ECG signal graph while comparing them with ideal
sample.
At first from ‘Select sample’ option ‘Sample 1’ was selected and
by pressing ‘PLAY’ button audio was played. Test subject
listened the sound and viewed the ECG signal in graph as shown
in Fig. 2(a). Then ‘Atrial Fibrillation’ button was pressed, which
played the audio sound and plotted ECG signal for Atrial
(e) Fibrillation sample and then test subject was asked whether the
Fig. 1 ECG samples used for sonification (a) normal ECG (b) Atrial selected sample matches ‘Atrial Fibrillation’ or not as shown in
Fibrillation (c) Myocardial Infarction (d) Myocarditis (e) Coronary Artery Fig 2(b). Like these all four diseases and ‘Healthy Man’ button
was checked whether they match the selected sample or not. And
Fig. 1 shows the ECG signal of ECG samples used in our likewise other four samples from pop-up menu were played and
work. Total 5 categories of samples have been used. One of them checked. Result from user study is described in result section.
is normal ECG sample other four are pathological samples.

2.2 Tools
Data sonification technique was implemented in Matlab
(R2016a, Mathworks). Sounds samples were played on a
Bluetooth speaker. We have used the source codes from
https://pub.uni-bielefeld.de/data/2908653 [10].

2.2 Sonification Technique


From the 12 channel ECG data we have used first six leads (I,
II, III, aVR, aVL, avF) and a sound note is assigned to each
channel.
We took the sample ECG data in 10 sec time range and
sampled the raw dataset in MATLAB. The voltage of each ECG
channel was mapped to the amplitude of the corresponding sound
signal. The input signal was frequency modulated and amplitude
(a)
modulated sequentially with the help of carrier signal (sinusoidal
signal).
D minor scale (i.e. 146.83 Hz, 174.61 Hz, 220.00 Hz, 293.67
Hz, 349.23 Hz, 440.00 Hz) is selected over two octaves for better
audio sound. We also added a fixed set of harmonics fk=k. f0 to
each channel with k = 3, 4, 5 and amplitude as 15%, 5% and 5%
of the fundamental frequency f0. To avoid spectral confusion with
the ECG channels 4–6, the 2nd harmonic f2 has been left out
intentionally, which are octave-shifted fundamentals of channels
1–3. The overall sonification is a continuous stream of six notes
playing simultaneously. [10]

2.2 Data Analysis


To test the feasibility of ECG sonification process and to (b)
analyse the data twenty undergraduate students among our Fig. 2 Graphical User Interface (GUI) for playing audio sound generated
classmate were selected. A graphical user interface (GUI) was from ECG sample (a) ‘Sample 1’ is played (b) comparison of ’Sample 1’
built using GUIDE tool in MATLAB to train them and to play the with ‘Atrial Fibrillation’.
audio sound generated from the ECG data.
3 Results TABLE II.
SAMPLE IDENTIFICATION RESULT FOR VISUALIZATION TECHNIQUE
TABLE I.
SAMPLE IDENTIFICATION RESULT FOR SONIFICATION TECHNIQUE S1 S2 S3 S4 S5
S1 S2 S3 S4 S5 ID1
ID1 ID2
ID2 ID3
ID3 ID4
ID4 ID5
ID5 ID6
ID7
ID6
ID8
ID7
ID9
ID8
ID10
ID9
ID11
ID10
ID12
ID11
ID13
ID12
ID14
ID13
ID15
ID14
ID16
ID15
ID17
ID16
ID18
ID17
ID19
ID18
ID20
ID19
ID20
Table II shows the user study result for ECG signal visualization,
Table I shows the user study result for sonification technique, wrong detections are marked as black coloured cell. In Table II,
wrong detections are marked as black coloured cell. In Table I, S1 S1 corresponds to Myocardial Infarction, S2, S3, S4 and S5
corresponds to Myocardial Infarction, S2, S3, S4 and S5 corresponds to Coronary Artery, Myocarditis, healthy man and
corresponds to Coronary Artery, Myocarditis, healthy man and Atrial Fibrillation respectively. Black shade corresponds to wrong
Atrial Fibrillation respectively. Black shade corresponds to wrong detection. Using ECG signal graph visualization, 88 times test
detection. Five samples identification for 20 test subject, means subject identified the correct pathologies. User’s success rate is
total 100 times audio sounds were played, 85 times test subject highest in ‘Myocarditis’ detection, 19 users among 20
identified the correct pathologies. It is also observed that user’s successfully detected the ‘Myocarditis’ ECG signal from graph.
success rate is lower in ‘Atrial Fibrillation’ and ‘Coronary An interesting result can be found from the comparison of these
Artery’, five users failed to identify these two diseases. User’s two table. In sonification technique most of users successfully
success rate is highest in ‘Healthy Man’ detection, 19 users detected ‘Healthy Man’ sample from random samples but in graph
among 20 successfully detected the ‘Healthy Man’ ECG signal by visualization technique this is not the highest, instead maximum
audio sound. success rate is for ‘Myocarditis’ disease.

4 Conclusions
In our research, we have demonstrated sonification technique on
ECG dataset and conducted user study for sonification technique
along with data visualization technique. In practical scenario ECG
data of two different patients having same disease might not be
same which complicated ECG diagnosis by data processing and
data visualization. In that case sonification has the superiority,
since human auditory system can find melodies even if it is mixed
with noisy signal. From the analysis of our user study, we propose
that if we use sonification technique along with visualization,
accuracy of cardiac pathologies diagnosis will be higher.
REFERENCES
[1] T. Hermann, A. Hunt, J. G. Neuhoff, “The sonification handbook” Berlin,
Germany: Logos Publishing House, (2011).
[2] G.I. Mihalaș, M. Andor, A. Tudor, S. Paralescu, “Potential Use of Sonification
for Scientific Data Representation”, Romanian J. Biophys., Vol. 28, pp. 45–57,
(2018).
[3] G. Dubus, R. Bresin, “A Systematic Review of Mapping Strategies for the
Sonification of Physical Quantities”, PLoS ONE, vol. 8, Dec., (2013).

[4] M. Ballora, B. Pennycook, P. C. Ivanov, L. Glass and A. L. Goldberger, “Heart


Rate Sonification: A New Approach to Medical Diagnosis”, Vol. 37, No. 1, pp.
41–46, (2004)

[5] A. L. A. Blanco, S. Grautoff, T. Hermann, “Heart Alert: Ecg Sonification for


Supporting the Detection and Diagnosis of St Segment Deviations”, ISon. pp.
1-7, Dec, (2016)

[6] D. Worrall, B. Thoshkahna, N. Degara, “Detecting Components of an Ecg


Signal for Sonification”, ICAD, June, (2014).

[7] M. B. Schreiber, W. Lesiński, Ł. Trzciałkowski, “Image Data Sonification in


Medicine”, JMIT, vol. 12, pp. 177-182, (2008)
[8] V. Avbelj, “Auditory Display of Biomedical Signals through a Sonic
Representation: ECG and EEG Sonification”, MIPRO, May, (2012).

[9] H.G. Kaper, E. Wiebel, S. Tipei, “Data Sonification and Sound Visualization”,
vol.1, pp.48-51, Jul/Aug., (1999).

[10] J. N. Kather, T. Hermann, Y. Bukschat, T. Kramer, L. R. Schad & F. G.


Zöllner, “Polyphonic sonification of electrocardiography signals for diagnosis
of cardiac pathologies”, Sci Rep. 44549, (2017).

You might also like