R. B. Dubey and Anju Dahiya/ Elixir Digital Processing 88 (2015) 36137-36155 36137
Introduction
The ultrasound imaging is the most popular and cost effective
imaging modality for treatment of kidney stone disease. Image
segmentation is the low level image processing technique which
aims to partition an image into regions such that each region
groups pixels sharing similar attributes. Segmentation of
nontrivial images is one of the most difficult tasks in image
processing. Segmentation accuracy determines the eventual
success or failure of computerized analysis procedures. For this
reason, considerable care should be taken to improve the
probability of accurate segmentation. Here we segment the
image and detect the stones present in kidney. Individual errors
may occur during the interpretation of ultrasound image by
an untrained sonographer, while taking dimensions. Thus, for
the purpose of avoiding the dependability to the sonographers’
expertise, some image processing operations are applied for
segmentation and detection of stone. We use the evolutionary
techniques for segmentation of the kidney stone ultrasound
images based on electromagnetism optimization algorithm and
harmony search algorithm for determining multilevel threshold
for image.
Segmenting the calculi from the kidney images is very
useful for the medical diagnosis in analyzing the patient’s
data. Considering the importance of the kidney stone
segmentation, many research works are developed with different
techniques to accomplish the kidney stone segmentation process.
This work focuses on the detection of kidney stone using
segmentation approaches which uses the ultrasound kidney
image as input. The ultrasound images are very challenging and
prone to speckle noise so images are pre-processed firstly using
the image restoration methods and then segmented by the
proposed segmentation approaches. Afterwards, the stone is
extracted and the area of the detected stone is compared to
calculate the error and accuracy of methods used [14, 16].
The remainder of this paper is organized as follows. Section
2 reviews the relevant previous literature and highlights the
research motivation. Section 3 presents the materials and data
set used. Section 4 summarizes the details of methodology.
Section 5 provides results and discussions. Conclusions are
drawn in Section 6.
Related Work
Kidney stone disease is very prevalent among Indians out of
every 1000 Indian 3 people are suffering from kidney stone. The
problem of stone can be due to various reasons such as food
habits, salts present in drinking water and it could be genetic.
Ultrasound imaging modality is one of popular method used by
specialist to diagnose it. The reason behind the wide use of
ultrasound images is because they are non-invasive, portable,
radiation free, and affordable. Segmentation helps to detect and
analyze the images which provide useful information regarding
the progress of the disease [14]. The presence of speckle noise
in ultrasound images is a problem which requires preprocessing
of images before segmentation [13, 14, 16]. Normally,
segmentation process is based on the image gray-level
histogram, namely image histogram thresholding. The threshold
based methods are parametric and non-parametric types. In
parametric approaches, it is necessary to find some parameters
like probability density function which models each class and
these approaches are time consuming and computationally
complex. Whereas the non-parametric involves use of several
terms like between class variance, entropy, error rate etc. which
is needed to be optimized to find an optimal threshold values.
For bi-level thresholding the classical methods are Otsu method
which chooses the optimal thresholds by maximizing the
between class variance of gray levels and Kapur’s method finds
the optimal threshold values by maximizing the entropy of
histogram. Multilevel thresholding uses a number of
thresholds in the histogram of the image to separate the
pixels of the objects in the image. Although both Otsu and
Kapur’s method can be expanded to multilevel thresholding but
the problem lies in the selection of the optimal thresholds due to
computational complexity which increases exponentially with
Tele: 9971833006
E-mail addresses: rbdubeyster@gmail.com
© 2015 Elixir All rights reserved
A Comparison of Two Kidney Stone Segmentation Techniques
R. B. Dubey and Anju Dahiya
Hindu College of Engg, Sonepat, Haryana, India.
ABSTRACT
Kidney stone disease is very common among Indians and approximately 5-7 million patients
suffer from stone disease. The most popular imaging modality for its treatment is ultrasound
imaging. Segmentation accuracy determines the eventual success or failure of computerized
analysis procedures. Traditional methods available for segmentation of kidney stones are not
accurate. We use the evolutionary techniques for segmentation of the kidney stone
ultrasound images for the first time based on electromagnetism optimization algorithm and
harmony search algorithm for determining multilevel threshold of images. In most cases,
the optimal thresholds are found by the minimizing or maximizing an objective function,
which depends on the positions of the thresholds. We identify a class of objective functions
for which the optimal thresholds can be found using algorithms with low time complexities.
The Kapur’s entropy function and Otsu’s between class variance is maximized using these
algorithms and an optimal threshold values at different levels are found out. The area of
extracted stone is calculated and compared with the area given by the expert radiologist. The
other metric used to compare the segmented image and the original image is the peak signal
to-noise-ratio (PSNR) value. The relative error is also calculated to show the segmentation
accuracy of each method.
© 2015 Elixir All rights reserved.
ARTICLE INFO
Article history:
Received: 22 September 2015;
Received in revised form:
28 October 2015;
Accepted: 03 November 2015;
Keywords
Kidney stone,
Segmentation,
Electromagnetism optimization
algorithm,
Harmony search algorithm.
Elixir Digital Processing 88 (2015) 36137-36155
Digital Processing
Available online at www.elixirpublishers.com (Elixir International Journal)