Content uploaded by Zainal Arief
Author content
All content in this area was uploaded by Zainal Arief on Jan 04, 2022
Content may be subject to copyright.
EMITTER International Journal of Engineering Technology
Vol.3, No.2, December 2015
ISSN: 2443-1168
Copyright © 2015 EMITTER International Journal of Engineering Technology - Published by EEPIS
99
Development of Healthcare Kiosk for Checking Heart Health
Riyanto Sigit, Zainal Arief, Mochamad Mobed Bachtiar
Politeknik Elektronika Negeri Surabaya
Kampus PENS, Jalan Raya ITS Sukolilo, Surabaya 60111
Tel: (031) 594 7280; Fax: (031) 594 6114
E-mail: riyanto@pens.ac.id,zar@pens.ac.id, mobed@pens.ac.id
Abstract
The main problem encountered nowadays in the health field,
especially in health care is the growing number of population and the
decreasing health facilities. In this regard, healthcare kiosk is used as an
alternative to the health care facilities. Heart disease is a dangerous one
which could threaten human life. Many people have died due to heart
disease and the surgery itself is still very expensive. To analyze heart
diseases, doctor usually takes a video of the heart movement using
ultrasound equipment to distinguish between normal and abnormal case.
The results of analysis vary depending on the accuracy and experience of
each doctor so it is difficult to determine the actual situation. Therefore, a
method using healthcare kiosk to check the heart health is needed to help
doctor and improve the health care facilities.
The aim of this research is to develop healthcare kiosk which can
be used to check the heart health. This research method is divided into
three main parts: firstly, preprocessing to clarify the quality of the
image.In this section, the writers propose a Median High Boost Filter
method which is a combined method of Median Filtering and High Boost
Filtering. Secondly, segmentation is used to obtain local cavities of the
heart. In this part, the writers propose using Triangle Equation that is a
new method to be developed. Thirdly, classification using Partial Monte
Carlo method and artificial neural network method; these methods are
used to measure the area of the heart cavity and discover the possibility of
cardiac abnormalities. Methods for detecting heart health are placed in
the kiosk. Therefore, it is expected to facilitate and improve the
healthcare facilities.
Keywords: Healthcare kiosk, heart health, reprocessing,
segmentation, classification.
1. INTRODUCTION
Heart disease is very dangerous and it causes the death of many people
throughout the world. Clinics have focused their research on diagnosing
cardiac disease. Cardiac function can be measured using ultrasound,
Magnetic Resonance Imaging (MRI), Computed Tomography (CT), X-Ray,
Volume 3, No. 2, December 2015
EMITTER International Journal of Engineering Technology, ISSN: 2443-1168
100
Positron Emission Tomography (PET) and Single Photon Emission
Tomography (SPECT).
This study used ultrasound and echocardiography to measure cardiac
function. Echocardiography is advantageous because it is affordable, widely
used, provides a great amount of information, non-invasive, does not expose
the patient to radiation and it can be applied to patients in critical condition.
The results are directly evident. Standard images used in echocardiography
are the short-axis, long-axis, two-chamber and four-chamber views. At
present, the reading of echocardiograms by experts is conducted using
standard clinical practice. However, this task requires accuracy and
experience in diagnosing the heart cavity structure.
Various studies have focused on segmenting cardiac cavities [1, 2].
However, there is still room for the innovation and development of new
methods and algorithms for further improvement. To perform segmentation
and detection of echocardiographic images, a number of researchers have
used short-axis images [3]. Klingler et al. [4], Lacerda et al. [5] and dos Reis et
al. [6] used semi-automatic detection methods, whereas Hae-Yeoun et al. [3]
and Ohyama et al. [7] developed fully automated detection methods. Works
on using snake’s algorithm to perform segmentation of cardiac cavity are
reported by Jierong, et al [8], Moursi&El-Sakka[9] and Ranganath[10].
In this study, to develop kiosk for checking heart health with an
alternative and implementation the simpler method called triangle method is
proposed to delineate the boundary of cardiac cavity from echocardiographic
images. In order to obtain an initial estimate of the boundary, combinations
of filtering and morphological operations are applied to the original image.
The filtering process involves application of high boost filter along with
morphological and thresholding operators to eliminate noise and convert the
image into a binary image. The use of a high boost filter is meant to enhance
the high frequency components without affecting the low frequency
components. This filter is followed by a negative Laplacian filter for edge
detection. A region filter is then applied to confine the boundary detection
region. The triangle method is then used to reconstruct a more precise
border by pruning and connecting the unwanted sections of the boundary
and lines. All the method will be implemented in the healthcare kiosk.
2. RELATED WORKS
The main objective of this work is to development of healthcare kiosk
for checking hearth health. In [1], Noble and Boukerroui survey several
methods and techniques for ultrasound segmentation. Petitjean and
Dacher[2] review segmentation methods in short axis cardiac MRI. Lee et al.
[3] used automatic segmentation for cardiac images using iterative
thresholding and an active contour model. Klingler et al. [4] applied
mathematical morphology to perform the segmentation of echocardiography
images. Lacerda et al. [5] presented radial search for the segmentation of
cardiac cavities. Chalana et al. [11] used multiple active contour models for
Volume 3, No. 2, December 2015
EMITTER International Journal of Engineering Technology, ISSN: 2443-1168
101
cardiac boundary detection. Laine and Zong[12] applied border identification
that depended on the shape modeling and border reconstruction from a set
of images. Ohyama et al. [7] presented a ternary threshold method for
detection of the left ventricular endocardium. Maria et al. [6] used semi-
automatic detection of the left ventricular border. Sigit et al. [13] applied the
triangle equation for automatic segmentation of the cardiac cavity, but only
worked for cases where there are only open major boundary without isolated
minor closed contours. In this paper, improvement of this method is
presented that is able to handle more general cases.
Segmentation using snakes have been conducted by several
researchers. Kass et al. [14] was the first to introduce the snake algorithm.
Chenyang and Prince[15] developed a new model of external forces for
snake, called gradient vector flow (GVF). Jierong et al. [8] used morphological
operations and the snake model to detect the boundary of echocardiographic
images. Moursi and El-Sakka[9] presented an efficient algorithm for
determining a good initial centroid contour in ultrasound images with the
snake algorithm. Ranganath[10] developed algorithms based on snakes for
extracting blood pool contours from cardiac MRI. Itai et al. [16] used
automatic detection of abnormal shadow by using snake for segmentation of
lung area. Jaffar et al. [17] describe a method for full automatic segmentation
based on Fuzzy entropy and morphological image processing techniques
from CT-Scan images.Sigit et al. [18] proposed a method for segmentation of
cardiac cavity image using search contour and snakes.
The snake is an active contour model. Snake can be used to find the
boundary of an object in an image starting from an initial boundary estimate.
It searches boundary images by minimizing the energy function. The total
energy of the snake is defined as in (1).
( ) ( ) ( )
(
)
dssvsvsvE
∫
−+=
λβα
22
"'
2
1
(1)
The weakness of the snake algorithm is that it is dependent on the
initial boundary estimate and setting of energy function variables (
λ
β
α
,, ).
In this study, we propose an algorithm called triangle method combined with
filtering and a morphology operation that is independent of initial estimate
and does not require any variable to be set.
3. ORIGINALITY
In this paper, development of healthcare kiosk for checking hearth
heath is proposed using triangle method.The main task of the proposed
triangle method is to prune and connect these boundaries to form a single
closed boundary to delineate the cardiac cavity. This is achieved by
successively scanning the points on noisy boundary using the corners of a
triangle subtended from the center of the initial boundary.
Triangle method has been successful gives better results compared
with the snake
algorithm
. The proposed method can provide solutions of
Volume 3, No. 2, December 2015
EMITTER International Journal of Engineering Technology, ISSN: 2443-1168
102
image segmentation for cardiac cavity images.The triangle method for image
segmentation is a novelty and not used mainly for segmentation.Results from
test using actual echocardiographic images show that the proposed method
give better accuracy compared to snake algorithm. The proposed method is
also easier to use as it does not need any tuning parameter.
4. SYSTEM DESIGN
The preprocessing steps involves several filtering processes such as
high boost, morphological, thresholding, negative Laplacian and region
filtering as shown in Figure 1.
Figure 1. Block diagram of the pre-processing steps
The procedure begins by reading the input file, i.e., the AVI video files
comprising the stored ultrasound images of the short-axis display of the
cardiac cavity. The input file is read by view image, which then converts the
video recording into a sequence of image frames to enable the images to be
processed and analyzed.
Next, a high boost filter is implemented to enhance the image because
the images obtained via view image are not clear. By using the high boost
filter, the image can be improved, and the cardiac cavity can be made more
prominent because in high boost filtering, the high frequency components of
the image are enhanced, while the low frequency components remain mostly
unchanged.
Though the image is somewhat improved, there is still noise due to
the enhancement of the high frequency components. To suppress this noise, a
morphological technique using the open and close operators is used to
remove speckle noise and sharpen the edges of the cavity, which is followed
by thresholding to convert the image into a binary format. Next, the binary
image is subjected to a negative Laplacian filter to determine the boundary of
the cavity. The implementation of a negative Laplacian filter involves taking
derivatives and performing convolution on the binary image to extract the
high-pass information directly. By doing so, a sharpened image can be
obtained, and the edges of the cavity can be clearly determined. Next, the
image is further subjected to the region filtering technique to eliminate small
and insignificant contours. Before region filtering, the center point and radius
of the circles are preset. After region filtering, the boundary of the cardiac
cavity in the image is almost evident but still contaminated with other
contours. To completely remove these other unwanted contours and
maintain only the desired contour of the cardiac cavity, a method based on
intersecting connecting line is used. The use of the intersecting connecting
was reported and described in [13].
High Boost
Filter
Filtered
Image
Morphology &
Thresholding Negative
Laplacian Region
Filter
Original
Image
Volume 3, No. 2, December 2015
EMITTER International Journal of Engineering Technology, ISSN: 2443-1168
103
4.1 Cardiac Image View
Typically, cardiac cavity images are presented using standard viewing
methods, such as the short-axis and long-axis views. In this research, short-
axis images are processed and analyzed to test the proposed segmentation
and detection algorithm using the triangle equations.
The procedure starts by reading the input file, i.e., the AVI video files
comprising the stored ultrasound images of the short-axis display of the
cardiac cavity. The input file is read using view image, which then converts
the video recording into a sequence of image frames to allow the images to be
processed and analyzed.
4.2 High Boost Filter
The high boost filter is implemented first. Itis one of the sharpening
operators in image processing. A high pass filter reduces the original image
using a low pass image, asshown in (2).
High pass = Original – Low pass (2)
A high boost filter multiplies the original image by a factor A, and the
result is then reduced by the low pass image, asshown in (3).
High boost = A * Original – Low pass (3)
Alternatively, the original image can be multiplied by (A–1), and the
result can be added to the high pass image, as shown in (4). In the case where
A is equal to1, it is called a high pass filter, but if A is greater than 1, it is
called a high boost filter.
High boost = (A–1) * Original + Original – Low pass
High boost = (A–1) * Original + High pass (4)
The high-boost filter is used to enhance high frequency component
while maintaining the low frequency components. The kernel from the high
boost filter is shown in (5):
W
highboost
= (A-1) * W
original
+ W
highpass
(5)
For example, W
highboost
= (A-1) *
000
010
000
+
−−− −
010
141
010
W
highboost
= (A-1) *
000
010
000
+
−−− −− −−−
111
181
111
Volume 3, No. 2, December 2015
EMITTER International Journal of Engineering Technology, ISSN: 2443-1168
104
Figure 2 and Figure 3 show the effect of the high pass filter and high
boost filter. Figure 2(a) is the short-axis image of the left ventricular cardiac
cavity, which was obtained from echocardiography. Figure2(b)is the effect of
a high pass filter with central kernels 4. Figure2(c) shows the effect of a high
pass filter with the center of kernel8. Figure 3 represents the effects of the
high boost filter with A=2, 4, 8, 16, 32 and 64. The image becomes sharper as
the factor A increases.
(a) (b) (c)
Figure 2. The effect of the high pass filter. (a) Original short axis image. (b) High
pass 4. (c) High pass 8.
(a) (b) (c)
Figure
3. The effect of the high boost filter. (a) Original short axis image. (b) High
pass 4. (c) High pass 8.
Sigit et al. [13] presented high boost filter with uses the spatial mask
from Lacerda[5]. Inthisstudy, an Avalue of 32 was used.Figure 4 shows the
spatialmaskassociated with this value of A. Using this high boost filter, the
image can be enhanced, and the cardiac cavity can be made more visible as
shown in Figure 3.
-1 -1 -1
-1 40 -1
-1 -1 -1
Figure 4. Spatial mask used for the high boost filter
4.3 Morphology and Thresholding
After implementing the high boost filter, the morphology operation is
used to remove noise. Basic morphology operations are dilation and erosion.
Both dilation and erosion are produced by the interaction of a set of
structureelements with the image. This structure element has the shape and
origin shown in Figure 5(a).
Suppose A is the set of pixels, and B is a structure element. Structure
element B dilates set A by filling holes with certain shapes and sizes provided
by the structure elements, such as in Figure 5(b). Structure element B erodes
Volume 3, No. 2, December 2015
EMITTER International Journal of Engineering Technology, ISSN: 2443-1168
105
set A by eliminating certain shapes and sizes of the structure provided by the
structure elements, such in as Figure 5(c).
(a) (b) (c)
Figure 5. Morphologyoperation. (a)Structureelements (b)Dilation (c)Erosion
Dilation
BA
⊕
can be written as in formula (6), while the erosion
BA
Θ
can be written as in formula (7), where
B
ˆ is the reflection of B followed by a
shift z.
(6)
(7)
This study uses morphological operations with the opening and closing
algorithm. The main purpose of the opening and closing algorithm is to
reduce speckle noise in the cardiac cavity image. The opening algorithm
involves eroding image A by B and dilating it by B. Mathematically, it is
achieved by using the mathematical notation of the opening algorithm shown
in Equation 8. The closing algorithm is implemented by dilating image A and
eroding it by B. The mathematical notation of the closing algorithm is shown
in Equation 9.
BBABA
⊕
Θ
=
)(o (8)
BBABA
Θ
⊕
=
•
)( (9)
Figure6 illustrates the effect ofmorphology using the structure
elements circle, crossandellipse. In this study, an elliptical structure element
with a 3x3 matrix structure was used because the shape and number of these
structures was found to give the best result.
Figure 6. The effect ofmorphology by using the structure lements circle, cross and
ellipse
}
ˆ
{ΦA)Bz|(BA
z
≠∩=⊕
}{ Az|(B)BA
z
⊆=Θ
Volume 3, No. 2, December 2015
EMITTER International Journal of Engineering Technology, ISSN: 2443-1168
106
4.4 Negative Laplacian Filter
Next, the negative Laplacian filter, which is a second derivative filter,
was realized to find areas of rapid change in the image (i.e., edges). Image
edges canbe obtainedin various ways, but the two most commonly used
methods for finding the edge of the image a research-based and zero-
crossing-based methods. The former method findsthe maximum and
minimum from the image edge based on the first derivative, and this method
is also called the gradient method. The latter method finds the image edge
using the second derivative, and this method is also known as the Laplacian.
Figure7 shows the filtered images using various edge detection gradient
methods.
(a)
(b)
(c)
(d)
(e)
(f)
Figure 7. Filtered images using different edge detection methods. (a) Original image.
(b) Robert. (c) Prewitt. (d) Sobel. (e) Canny. (f) negative Laplacian.
It appears that the Canny and negative Laplacian methods give good
results but differ in terms of computational speed. Inthis paper, the negative
Laplacian for edge detection process was used because it is faster than the
Canny method. The Canny method requires a longer processing time than the
Laplacian method because it requires the use of the Gaussian derivative and
two threshold values to detect both strong and weak edges. There are several
ways to find an approximate discrete convolution kernel that approximate
the effect of the Laplacian. A possible kernel is shown in Figure 8 below:
0 1 0
1 -4 1
0 1 0
Figure 8. Kernel used for negative Laplacian
Volume 3, No. 2, December 2015
EMITTER International Journal of Engineering Technology, ISSN: 2443-1168
107
4.5 Region Filter
Subsequently, the region filter was used to eliminate small contours.
The region filter scans the contour and calculates the area of each contour.
Regions with an area that is smaller than the pre-determined threshold are
eliminated from the contour [5]. The threshold value was set to 25 pixels and
was empirically determined. Figure 9(b) shows the effect of applying region
filter to the image given in Figure 9(a).
(a) (b)
Figure 9. (a) Imaged formed after high boost filtering and morphology operation.
(b) Image after region filtering.
4.6 Removing remote contours
The final process in the pre-processing step is to remove contours that
are located far from the major boundary. This is achieved by finding the
centroids of all existing contours using Equation 10, as shown below.
=
∑∑
==
n
yk
n
xk
ccentroid
n
k
n
k11
,)(
(10)
A connecting line is then formed from the center of the boundary to the
centroids of each contour. The contour will be maintained if the respective
connecting line intersects the inner boundary, and will be removed
otherwise. Figure 10 show the image after applying this step.
Figure 10. Image after applying intersecting connecting line criteria.
Volume 3, No. 2, December 2015
EMITTER International Journal of Engineering Technology, ISSN: 2443-1168
108
4.7 The Triangle Method
Figure 11(a) shows a triangle, where A, B, and C are the corners and a,
b, and c are the distances between the corners. If a, b, and c are known, then
the angle of the corner can be calculated using Equations 11 and 12.
a
2
= b
2
+ c
2
- 2bc (cos A) (11)
A = acos(( b
2
+ c
2
- a
2
) / 2bc) (12)
The previous process may produce an image with open and closed
boundaries. The boundary is closed when all points are connected, as shown
in Figure 11(a). The boundary is open if there are disconnected points or
endpoints, as shown in Figure 11(b).
(a) (b)
Figure 11. Boundary of cardiac cavity (a) close boundary (b) open boundary
The boundary of the cardiac cavity can be obtained directly if the contour
is
closed, but if the boundary of the cardiac cavity is open, it can be obtained by
using the triangle method according to the following steps.
1. The first step is to determine the approximate center of the region that
seems to be enclosed by the boundary, which can be done by calculating
the centroid as in Equation 9. This point is denoted as point A in Figure
12(b).
2. The two endpoints of the boundary are located and are marked as points B
and C in the same figure. The triangle BAC is formed from these three
points with sides a, b and c.
3. The gap in the boundary is minimized by first fixing one of the end points,
such as point B in Figure 12(c), and allowing point C to transverse inward
along the boundary until the point at which the angle BAC is minimized is
found. A similar action is then performed for point B to obtain the result
shown in Figure 12(d).
4. The open boundary is then closed by connecting a line from point B to
point C.
Volume 3, No. 2, December 2015
EMITTER International Journal of Engineering Technology, ISSN: 2443-1168
109
(a) (b) (c) (d)
Figure 12.(a) Triangle sides and angles (b) triangle on the cardiac cavity image. (c)
creating a small corner (d) example of finding boundary
The opening gap of the open boundary can be made as small as possible
by searching for the two points from each end point such that the angle
subtended from these points from the center is minimized. The search starts
from one end point of the open boundary and moves inside until it traverses
1/3 of the length of the open boundary. Figure 13(a) shows the variation of
the angle of the triangle BAC with B fixed and point C moving inward. The
plot shows that the minimum angle is 32.3˚at point 52 from the starting point
C. Similarly, with point C fixed at the new location, point B moves inside, and
the minimum angle is at point 28, as shown in Figure 13(b)
(a)
0
20
40
60
80
100
120
140
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85
Angle BAC (degree)
Distance of C from starting position (pixel)
Volume 3, No. 2, December 2015
EMITTER International Journal of Engineering Technology, ISSN: 2443-1168
110
(b)
Figure 13. The angle value is calculated at the two end points. (a) Point C moving
inside with point B fixed. (b) Point B moving inside with point C fixed at new
location.
There will also be cases when there is an open major boundary with
isolated minor closed contours, as shown in Figure 14(a).
(a) (b) (c) (d)
Figure 14. Using the Triangle Equation method to close the open boundary
For such cases, the following procedure is proposed to form the closed
boundary.
1. Minimize the gap in the open boundary using Steps 1-3 in the procedure
described earlier. Using Figure 14(a) as an example, the initial triangle was
formed, as shown in Figure 14(b). The opening in this boundary is then
minimized, as shown in Figure 14(c). The boundary formed after this
process is shown in Figure 14(d).
2. The minor closed contours are then merged with the major boundary
using the follow steps:
a) Determine the centers of the close contours by calculating the average
position of point x and the average position of point y on the contour by
using Equation 9.
b) Locate the end points of the open boundary. These centers and end points
are called the anchor points. Figure 15(a) shows the locations of these
0
20
40
60
80
100
120
140
1
4
7
10
13
16
19
22
25
28
31
34
37
40
43
46
49
52
55
58
61
64
67
70
Angle BAC (degree)
Distance of B from starting position (pixel)
Volume 3, No. 2, December 2015
EMITTER International Journal of Engineering Technology, ISSN: 2443-1168
111
anchor points.
c) Find the points in the adjacent contours that are closest to the anchor
points. These points are called the nearest points.
d) Starting with one of the end points, connect it to the nearest point. From
this point, take the path along the contour that is closer to the center of the
main boundary until another nearest point is reached. From this point,
make a connection to the nearest point in the neighboring contours. The
process continues until it is connected to another end point, as shown in
Figure 15(b).
e) In the previous step, the sections of the contour between two nearest
points were retained as part of the boundary. The other section is then
removed to give the final boundary, as shown in Figure 15(c).
(a) (b) (c)
Figure 15. Constructing the boundary of the cardiac cavity with the open and some
closed contours
5. EXPERIMENT AND ANALYSIS
The proposed triangle algorithm was tested using a dataset composed
of actual ultrasound (US) videos from 109 patients. Seventy-eight videos
contained good quality images, and the rest were of poor quality and
required additional image enhancement and processing techniques. The
testing datasets were extracted from these video recordings, and each
comprised 6 to 10 repeated frames. The typical image size was 320 pixels
wide and 240 pixels high. Figure 16 shows the delineated cardiac cavity for
several video frames from diastole to systole.
Figure 16. Delineated cardiac cavity during different stages of cardiac cycle
The information about the movement of the boundary of the cardiac
cavity provides useful information for the physician on the condition of the
myocardium. Useful information that can be extracted from the boundary is
Volume 3, No. 2, December 2015
EMITTER International Journal of Engineering Technology, ISSN: 2443-1168
112
the area of the cavity that can be used to infer the cardiac pumping
performance. In order to assess the performance of the proposed method, the
areas of the cardiac cavity in the echocardiographic images from different
stages of the cardiac cycle are computed using the triangle method and snake
algorithm. These areas are then compared to areas of the cavity delineated
manually by the physician. The unit for the area is the number of pixels in the
boundary. The comparison is presented in Table 1. From this table, we can
see that the areas of the cardiac cavity obtained using the triangle method
have an average error of 6.56, compared to an average error of 13.68 based
on snake algorithm.
Table 1. Comparison of cardiac cavity areas using triangle and snake
Images
(Frame)
Area based
on manual
Area based
on snake
algorithm
Area using
triangle
method
Error using
snake
algorithm
Error using
triangle
method
1
2
3
4
5
6
7
8
9
4943
4669
4481
3955
3176
2731
1901
2236
2013
5988
5735
4800
4191
4059
3023
2087
2104
2252
5329
4995
4673
4038
3497
2515
1817
2355
2217
21.14
22.83
7.12
5.97
27.80
10.69
9.78
5.90
11.87
7.81
6.98
4.28
2.10
10.11
7.91
4.42
5.32
10.13
Average error 13.68 6.56
Figure 17 shows that the calculated areas of the triangle method are
almost the same as those computed from the boundary delineated manually
by the physician from frame 1 to frame 9.The triangle method clearly shows a
better accuracy than the snake algorithm.
Figure 17. Comparison of cardiac cavity areas using several methods
0
1000
2000
3000
4000
5000
6000
7000
1 2 3 4 5 6 7 8 9
Area of boundary (pixel)
Frame
Manual
Snake
Triangle
Volume 3, No. 2, December 2015
EMITTER International Journal of Engineering Technology, ISSN: 2443-1168
113
6. CONCLUSION
The proposed method presents solution for healthcare kiosk for
checking heart health. The proposed method can made preprocessing image
to be clearly, segmentation can get the heart cavity region and the
classification method can detect heart abnormalities.
Acknowledgements
The authors would like to thank the Ministry of Research Technology
and High Education and the Indonesia government for funding this research
through research grant and also would like to thank the Politeknik
Elektronika Negeri Surabaya.
REFERENCES
[1] J. A. Noble and D. Boukerroui, Ultrasound image segmentation: a
survey,IEEE Transactions onMedical Imaging,vol. 25, pp. 987-1010,
2006.
[2] C. Petitjean and J.-N. Dacher, A review of segmentation methods in
short axis cardiac MR images, Medical Image Analysis, vol. 15, pp. 169-
184, 2011.
[3] L. Hae-Yeoun, N. C. F. Codella, M. D. Cham, J. W. Weinsaft, and W. Yi,
Automatic Left Ventricle Segmentation Using Iterative
Thresholding and an Active Contour Model With Adaptation on
Short-Axis Cardiac MRI, IEEE Transactions on Biomedical Engineering,
vol. 57, pp. 905-913, 2010.
[4] J. W. Klingler, Jr., C. L. Vaughan, T. D. Fraker, Jr., and L. T. Andrews,
Segmentation of echocardiographic images using mathematical
morphology, IEEE Transactions onBiomedical Engineering, vol. 35, pp.
925-934, 1988.
[5] S. G. Lacerda, A. F. da Rocha, D. F. Vasconcelos, J. L. A. de Carvalho, I. G.
Sene, and J. F. Camapum, Left ventricle segmentation in
echocardiography using a radial-search-based image processing
algorithm, Medicine and Biology Societyin Engineering, EMBS 2008.
30th Annual International Conference of the IEEE, pp. 222-225, 2008.
[6] M. d. C. dos Reis, A. F. da Rocha, D. F. Vasconcelos, B. L. M. Espinoza, F. A.
d. O. Nascimento, J. L. A. de Carvalho, S. Salomoni, and J. F. Camapum,
Semi-automatic detection of the left ventricular border, Medicine
and Biology Societyin Engineering,EMBS 2008. 30th Annual International
Conference of the IEEE, pp. 218-221, 2008.
[7] W. Ohyama, T. Wakabayashi, F. Kimura, S. Tsuruoka, and K. Sekioka,
Automatic left ventricular endocardium detection in
echocardiograms based on ternary thresholding method,
Proceedings 15th International Conference onPattern Recognition, pp.
320-323 vol.4, 2000.
Volume 3, No. 2, December 2015
EMITTER International Journal of Engineering Technology, ISSN: 2443-1168
114
[8] C. Jierong, F. Say Wei, and S. M. Krishnan, Watershed-presegmented
snake for boundary detection and tracking of left ventricle in
echocardiographic images, IEEE Transactions onInformation
Technology in Biomedicine,vol. 10, pp. 414-416, 2006.
[9] S. G. Moursi and M. R. El-Sakka, Initial Contour for Ultrasound
Carotid Artery Snakes, IEEE International Symposium onSignal
Processing and Information Technology, pp. 390-395, 2007.
[10] S. Ranganath, Contour extraction from cardiac MRI studies using
snakes, IEEE Transactions onMedical Imaging,vol. 14, pp. 328-338,
1995.
[11] V. Chalana, D. T. Linker, D. R. Haynor, and K. Yongmin, A multiple
active contour model for cardiac boundary detection on
echocardiographic sequences, IEEE Transactions onMedical
Imaging,vol. 15, pp. 290-298, 1996.
[12] A. Laine and Z. Xuli, Border identification of echocardiograms via
multiscale edge detection and shape modeling, Proceedings
International Conference onImage Processing, 1996, pp. 287-290 vol.3,
1996.
[13] R. Sigit, M. M. Mustafa, A. Hussain, O. Maskon, and I. F. Nor, On the use
of collinear and triangle equation for automatic segmentation and
boundary detection of cardiac cavity images, Adv Exp Med Biol, vol.
696, pp. 481-8, 2011.
[14] M. Kass, A. Witkin, and D. Terzopoulos, Snakes: Active contour
models, International Journal of Computer Vision, vol. 1, pp. 321-331,
1988.
[15] X. Chenyang and J. L. Prince, Snakes, shapes, and gradient vector
flow, IEEE Transactions onImage Processing,vol. 7, pp. 359-369, 1998.
[16] Y. Itai, H. Kim, and S. Ishikawa, A segmentation method of lung areas
by using snakes and automatic detection of abnormal shadow on
the areas, International Journal of Innovative Computing, Information
and Control, vol. 3, pp. 277-284, 2007.
[17] M. A. Jaffar, A. Hussain, A. M. Mirza, and A. u. Ch, Fuzzy Entropy and
Morphology based fully automated Segmentation of Lungs from CT
Scan Images, International Journal of Innovative Computing,
Information and Control vol. 5, pp. 4993–5002, 2009.
[18] R. Sigit, M. M. Mustafa, A. Hussain, O. Maskon, and I. F. Mohd Nor,
Cardiac Cavity Segmentation in Echocardiography using Search
Contour and Snakes, in Industrial Electronics Seminar 2009, Surabaya,
Indonesia, pp. C12-16, 2009.
