AN INTELLIGENT PHONOCARDIOGRAM FOR AORTIC STENOSIS DIAGNOSIS

aroocarmineAI and Robotics

Oct 29, 2013 (3 years and 11 months ago)

107 views

AN
INTELLIGENT PHONOCARDIOGRAM FOR AORTIC STENOSIS DIAGNOSIS


Arash Ghare
Baghi
1
,

Birgitta

Janerot
-
Sjoberg
2
, Peter Hult,
Per Ask
1

1

Physiological measurement, Department of biomedical engineering, Link
öp
ing University, Linköping, Sweden

2 Department
s

of
C
l
inical
S
cience,
I
ntervention and
T
echnology, Karolinska
Institutet & Clinical Physiology, Karolinska University
Hospital, Stockholm, Sweden


1.

Introduction


Cardiovascular diseases” is still the main factor of human mortality. A large percentage (as many a
s 70%) of the paediatric
referrals to the hospital for cardiac examinations, have normal hearts while to a lesser extent a number of pathological
patients are neglected by the screening procedures in primary healthcare centres [1].
The intelligent phonocar
diogram is a
non
-
invasive
, powerful

and inexpensive approach which can serve as a screening tool in hands of the practitioners

or nurses

in primary healthcare centres to select
appropriate

patients to undergo echocardiography which is by far a more expensi
ve
approach. Studies showed that a computer assisted phonocardiography will substantially improve the screening accuracy

[2]
.
Heart has a rhythmic mechanical action by which blood is pumped through the circulatory system. The mechanical action
constituted
from the heart muscle movements and
thereby
opening and

closure of
heart valves.
Phonocardiogram is an
acoustical recording of the heart sounds associated with the mechanical action. Auscultation of these sounds is still the mai
n
technique for screening he
art disease in the primary healthcare centres. However, the accuracy of this technique is influenced
by
physiological

restrictions exist in the human auditory system as well as other psychological and environmental effects.
Intelligent phonocardiogram is i
ndeed a computer based stethoscope, supported by the intelligent algorithms to
analyse heart
sound signal in order to
detect

pathological symptoms of the heart sound
. As the aforementioned negative effects are
efficiently attenuated thanks to the rapid pro
gresses in electronics systems and artificial intelligence, the undesirable factors
would be decreased in favour of the intelligent stethoscope. This device

help
s

the practitioners or nurses
in the primary
healthcare centres
as a decision support
or even a
s a screening tool
. The
powerful
algorithms could be incorporated into
smart phones or tablets for automatic diagnosis to be used as a home healthcare system.


2.

Method

Phonocardiogram signals were collected from the patients diagnoses as normal patients (30

cases) and the patients with aortic
stenosis (30 cases) according to the echocardiography. An electronic stethoscope which is commercially available was
utiliz
ed for signal recording.
Data recording was done according to the guidelines of the correspondin
g hospitals which were
in compliance with the declaration of Helsinki. Patients gave their informed consents prior to the data recording. The study
was approved by the corresponding local committees. In

this study
,
signal processing techniques in conjuncti
on with our
innovative pattern recognition methods are employed to screen and diagnose Aortic stenosis in children a
nd elderly people,
respectively. In order to evaluate the screening performance, random repeated sub sampling
(RRSS)
method is
applied to
es
timate an expected value

for the three performance measures; classification rate, sensitivity and specificity.


3.

Results

Balanced RRSS with 50%/50% of training/testing data is employed to evaluate the performance of the system.
The
average
value
s

for
accur
acy
, sensitivity and specificity after
5
0 iteration
s

are
estimated
as
88%
,
86% and 89%

respectively.

Regression analysis
confirme
d the significance of the diagnosis with R
>0.81.


4.

Discussion

Results showed that the proposed system can act as an
efficient

to
ol for screening and diagnosis of aortic stenosis. It is
evidenced by the previous studies that the screening accuracy of a specialized cardiologist who relies on auscultation is abo
ut
80%

[2]
, while the corresponding value of

the intelligent phonocardiogr
am

described is 88%
. As a consequence, the
performance of the system in screening is
at least as good as

a

specialized cardiologist

and
may
,

when tuned to a

l
arger
database including greater variety of different heart condition
s
,


it
has potential to furth
er improve
frugal
performance

availability

e.g. in smartphones
.


References


[1] C. G. DeGroff, S. Bhatikar, J. Hertzberg, R. Shandas, L. Valdes
-
Cruz and R. L. Mahajan, "Artificial neural network
-
based methods of screening heart murmurs in children," Circu
lation, vol. 103, pp. 2711
-
2716, 2001.

[2] R. L. Watrous, R. W. Thompson and S. J. Ackeman, "The Impact of Computer
-
assisted Auscultation on Physician
Referrals Asymptomatic Patients with Heart Murmurs," Clinical Cardiology, vol. 31, pp. 79
-
83, 2008.