Uploaded by Rainier Carlo Perez

Leukemia Detection

advertisement
See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/335857756
Segmentation and Detection of Acute Leukemia Using Image Processing and
Machine Learning Techniques: A Review
Article in AUS · September 2019
DOI: 10.4206/aus.2019.n26.2.60
CITATIONS
READS
0
648
3 authors:
Saif S Aljaboriy
Nilam Nur Amir Sjarif
Universiti Teknologi Malaysia
Universiti Teknologi Malaysia
1 PUBLICATION 0 CITATIONS
18 PUBLICATIONS 42 CITATIONS
SEE PROFILE
SEE PROFILE
Suriayati Chuprat
Universiti Teknologi Malaysia
78 PUBLICATIONS 374 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Real time ETL system View project
Segmentation and Detection of Acute Leukemia Using Image Processing and Machine Learning Techniques: A Review View project
All content following this page was uploaded by Saif S Aljaboriy on 17 September 2019.
The user has requested enhancement of the downloaded file.
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
๏‚จ
๏‚จ
Recepción/ 27 junio 2019
Aceptación/ 25 agosto 2019
Segmentation and Detection of Acute Leukemia Using
Image Processing and Machine Learning Techniques: A
Review
Segmentación y detección de leucemia aguda mediante
técnicas de procesamiento de imágenes y aprendizaje
automático: una revisión
Saif S Al-jaboriy1,2,
Nilam Nur Amir Sjarif1,
Suriayati Chuprat1
Razak Faculty of Technology and
Informatics, Universiti Teknologi Malaysia,
Malaysia.
1,2
Razak Faculty of Technology and
Informatics, Universiti Teknologi Malaysia,
Malaysia; and also with Al-Musiab
Technical College, Al-Furat Al-awsat
Technical University, Babylon, Iraq.
(e-mail: Ssaiff198@yahoo.com ;
com.syf4@atu.edu.iq)
ABSTRACT/ Leukemia is a blood cancer that affects the white blood cells in the bone marrow. The detection of acute leukemia
is dependent on counting the white blood cells percentages in the peripheral blood. This kind of cancer can be a fatal disease
if left without early detection. In practice, the manual microscopic examination methods are used for detection of acute
leukemia. But these manual methods are inaccurate, prone to errors, and time-consuming due to human factors such as
fatigue, stress, and lack of experience. Therefore, several of image processing techniques have been proposed to replace with
the manual methods. The results of this study: various computer-aided systems for acute leukemia diagnoses are reviewed;
these systems are including image acquisition, pre-processing, segmentation, extraction of feature, and classification. Also this
paper provides the pros and cons of existing computer-aided methods with their accuracy.
Key Words: Leukemia, image Acquisition, pre-processing, segmentation techniques, feature extraction Classification.
ARTÍCULO
RESUMEN / La leucemia es un cáncer de sangre que afecta los glóbulos blancos en la médula ósea. La detección de leucemia aguda
depende de contar los porcentajes de glóbulos blancos en la sangre periférica. Este tipo de cáncer puede ser una enfermedad mortal si
se deja sin detección temprana. En la práctica, los métodos de examen microscópico manual se utilizan para la detección de leucemia
aguda. Pero estos métodos manuales son inexactos, propensos a errores y requieren mucho tiempo debido a factores humanos como la
fatiga, el estrés y la falta de experiencia. Por lo tanto, se han propuesto varias técnicas de procesamiento de imágenes para reemplazar
con los métodos manuales. Los resultados de este estudio: se revisan varios sistemas asistidos por computadora para el diagnóstico de
leucemia aguda; Estos sistemas incluyen adquisición de imágenes, preprocesamiento, segmentación, extracción de características y
clasificación. Además, este documento proporciona las ventajas y desventajas de los métodos asistidos por computadora existentes con
su precisión.
Palabras clave: leucemia, adquisición de imágenes, preprocesamiento, técnicas de segmentación, extracción de características
Clasificación
511
1. Introduction
The human body contains several parts that
complement each other in the performance of
life functions. Blood is one of the most critical
parts of the human body that helps to pass
minerals and oxygen to all parts of the body.
It also maintains body temperature; as well as
protects
the
body
through
antibody
production. There are three essential
components in blood, namely white blood cells
(WBC), red blood cells (RBC), and platelets
[1]. Blood can be infected with various
diseases that lead to body dysfunction and
thus affect the life and health of humans.
Some of these diseases are not very serious
and are treatable such as anemia, and others
are very serious and not easy to treat such as
leukemia. Leukemia is a type of blood cancer
that attacks the bone marrow, which is
characterized by an increase in the number of
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
widely used for diagnosing the existence of
acute leukemia. Essentially, the microscope
images involve both normal and abnormal
blood cells images. Despite extensive
research,
the
accurate
detection
and
classification of acute leukemia by using
image-processing
techniques
remains
challenging. The reasons are due to the
variation of size, shapes, locations, and image
intensities for various kinds of acute leukemia
[19,20]. The detecting of leukemia using
image processing and computer vision
techniques consists of five stages, which
include image acquisition, pre-processing
image,
WBCs
segmentation,
feature
extraction, and classification of leukemia cells.
Fig 1 shows the stages of detection of
leukemia [21].
Fig1. Illustrate the stages of detection
leukemia
This aim of this study is to presents a detailed
review of computer systems that used to
detect and diagnosis acute leukemia (AL). This
paper is organized as follows: describes the
related work that is divided into five steps with
description each method followed by the
literature survey about leukemia. Then, the
techniques used for segmenting and detecting
acute leukemia will also be discussed.
2. Related research
Various researchers have proposed many
techniques in the field of medical image
processing. The majority of the researchers'
work is to find an optimal technique to help
the hematologists to detect acute leukemia
early. The major steps for diagnosis of acute
leukemia using image processing and machine
learning techniques involves pre-processing of
the medical images, segmentation (extract
the blast cells from the rest of image), feature
extraction, and then classification. These
steps are interrelated to each other, and the
success of the step will reflect on other steps.
In this section, we provide a survey of various
methods that have proposed for develop an
automatic system that used to detection of
acute leukemia cell by image processing.
2.1
Image acquisition
The detection of acute leukemia disease
through microscopic medical blood image
begins with the peripheral blood images
acquisition phase. At this stage, the medical
ARTÍCULO
immature WBCs named ‘blasts cells’.
Leukemia can be classified into two types
which are acute leukemia and chronic
leukemia [2]. This research focuses on the
acute type of leukemia. In acute leukemia, the
abnormal WBCs called immature blasts cells
malfunction, where it grows very quickly, and
the condition of the patient becomes very
dangerous if it not controlled quickly. Acute
leukemia infects the bone marrow and
progresses very fast. This kind of cancer can
infect children and adults too; as this disease
can develop many abnormal WBCs in their
body [3,4]. Acute leukemia can be classified
into
two
types
that
include
acute
lymphoblastic leukemia (ALL) and acute
myeloid leukemia (AML) based on a French–
American–British model. ALL is one of the
acute leukemia types that negatively affect
the lymphocytes (WBCs that fight infection).
The bone marrow produces a number of blasts
cells that develop quickly to be lymphocytes
[5]. This condition is fatal if left untreated
because it quickly spreads; the early diagnosis
of ALL is essential for saving patients’ life and
for recovery [6-8]. ALL occurs when the white
blood cells have yet to completely mature.
While in the case of AML, the immature white
blood cells are unable to fight infections
[11,12]. The detection of acute leukemia in
different types is essential for early diagnosis
and drug discovery [13,14]. The diagnosis of
acute leukemia is dependent on counting the
percentages of blasts cells in the blood.
Manual microscopic examination of the blood
smear is less accurate and time-consuming
[9,15,16]. Acute leukemia diagnosis requires
tools and automated solutions, and the ability
for early detection [17]. Therefore, the image
processing and computer vision systems can
play an important role in diagnosis of such
types of these medical problems. In this case,
the system uses to examine the blood sample
automatically in order to overcome the
limitations of manual examination. Can be
built these type of systems with use the
microscopic images to recognize types of
acute leukemia cells. Using these types of
systems will reduce the time and effort and
increase the accuracy of the diagnosis [10].
The detection and classification of leukemia
depend on color, texture, size, and shape of
WBCs image [18].
Medical images provide useful information
about normal and abnormal blood cells.
Presently, the microscope images are most
512
ARTÍCULO
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
513
blood images samples are captured from the
bone marrow by a camera to obtain digital
data that is used to test a proposed technique.
The light is one of the factors that has an effect
on the blood images, where the light
determines the quality and clarity of the
images. Thus, the pre-processing of the
images consists of the variety of techniques
that are used to change the domain of images
to another domain (this will be described in
detail in the next section) to be more easily
handled by the proposed systems [18,19].
Many researchers tested and evaluated their
systems with a few samples of blood images,
which are not available publicly. Therefore,
they could not compare their findings with
other researchers’ results. Hence, many
researchers used a public dataset that is
available online such as ALL-IDB1 (These kind
of samples were captured and collected by
experts at the M. Tettamanti Research Centre
in Monza, Italy).This is an example of section.
2.2 Pre-processing image
Enhancing medical images is an important and
complex step in computer vision, machine
learning and image processing techniques
because it used to improve the quality of an
image. Many medical images types (e.g. MRI,
microscope image, ultrasound Image, Nuclear
medicine, and others) are used to develop
diagnostic systems [20]. Many authors have
enhanced the blood image by converting to
another domain such as convert RGB to HSV
or to HSL domain to highlight the features of
objects for efficiently detecting an interesting
area [21]. Many factors can affect the quality
of images such as low contrast, false
background, and pepper noise. These kinds of
noises may occur because of the camera and
light condition during the capture of the
images. Currently, various techniques have
been proposed and suggested by Authors to
detect and enhance these kinds of factors in
order to make the blood images suitable for
segmenting region of interest (ROI) [22].
Histogram equalization is one of the
techniques that used for adjust the images
contrast. This technique is suitable to improve
and enhance the dark background and blood
image contrast [6,73]. Another technique is
linear contrast that also used to improve and
enhance the quality of the blood images by
increasing the contrast; this technique also
known as normalization of an image [7, 8]. In
addition, many proposed techniques are used
the minimum filter to highlight the lighter
object, which can easily recognize during the
segmentation process [10,72]. All the
technique that have been mentioned contain
disadvantages such as vulnerable to noise,
useful with image has low contrast, cannot
perform with blood images that having high
percentage of noise, i.e. led to remove some
detail from the image, and so forth. Therefore,
researchers need to find and propose new
techniques for enhancing image quality.
2.3
Segmentation image
Understanding the digital images or extracting
the information from some regions in the
image is one of the most crucial tasks in image
processing techniques. The first step in
detecting an object in the digital image is
image segmentation. In practice, image
segmentation is one of the important bases for
the recognition of the region of interest in the
image, and it is considered a hotspot in the
image processing, computer vision and
machine learning techniques [5,20,71].
Fig 2: WBCs segmentation methods.
The segmentation stage has divided an image
into a number of objects that have the same
features based on certain criteria in order to
extract the area of interest. Many authors
have been proposed different techniques to
segment blast cells (AML and ALL) from the
rest of the blood image as depicted in Figure
2; but the results of these techniques could
not produce ideal segmentation for the
complex blood cell images. In addition, many
factors made the segmentation of WBCs from
the rest of image a challenging process (e.g.,
the light, contrast, and quality of medical
images) [22,70]. Generally, different features
in the medical blood image can be used at
segmentation stage such as color, shape, and
texture and level intensity.
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
………… (1)
Where the ๐ผ is the original image in grey mode
and the ๐‘– is the image pixel value. And the ๐ผ' is
conceder as binary image.
The advantage of this method is the operation
speed is faster, and the calculation is simple;
especially, when the target objects have high
contrast. Nevertheless, the disadvantage of
this method is it is difficult getting the accurate
results; especially when there is a large
overlap of the gray-scale or no significant
gray-scale difference in the image. It is also
sensitive to gray-scale unevenness and noise
in the image. Therefore, it needed to be
combined with other methods to get better
performance [26].
2.3.2 Regional growth
The regional growth method is one of the
conventional region segmentation algorithms
that segment an image based on the similar
properties of the pixels in order to get the
target objects [27]. This method requires two
main steps that include selecting a seed pixel
and merging all the pixels that have similar
properties, and they are located around the
seed pixel. Its basic formula is as follows:
๐‘›
a) ๐‘ˆ๐‘–=1
๐‘…๐‘– = ๐‘….
b) ๐‘…๐‘– is a connected region, i = 1, 2, 3, . . . , n
c) ๐‘…๐‘– ∩ ๐‘…๐‘— = ๐œ™ for all ๐‘– = 1, 2, . . . , ๐‘›.
d) (๐‘…๐‘–) = TRUE for ๐‘– = 1, 2, . . . , ๐‘›.
e) (๐‘…๐‘– ∪๐‘…๐‘—) = FALSE for regions of ๐‘…๐‘– and ๐‘…๐‘—.
(a) Every pixel is part of the same region in
the image. While in (b) all the pixels should be
connected to each other in the region; (c) all
the region disjoint each other. The (d) are the
properties that are true or false with some
criteria. While in (e) all regions different with
predicate ๐‘ƒ
2.3.3 Clustering
In clustering method, there is a difference
between the pixel and the cluster center in the
blood image that depends on the density,
location, or other factors such as the K-means
clustering algorithm that is used to divide a
digital image according to the distance. The
process of implementing the K-means
clustering technique is expressed as follows:
(1) Randomly select the initial K clustering
centers value, (2) Calculate the distance from
each cluster center to each sample to return
the sample to the nearest center, (3) For each
cluster, the mean of the samples is connecting
with the new clustering centers [28]. The
advantage of this method is the algorithm is
simple and fast, and it is effective when used
with large datasets. In addition, it is close to
linear with the time, and thus it is suitable for
large-scale data sets. Mathematically it can be
defined as:
Where each of the ๐‘ฅ๐‘ will be assigned to one
(๐‘ก).While the centroid of the calculate the new
mean in the cluster. It can be defined as
following:
The disadvantage of the clustering method is
it is difficult to estimate the K clustering
number; where there is has no explicit
selection criteria [28]. Secondly, this method
is currently very expensive because of the
iteration. Finally, the clustering algorithm is a
partitioning method that depends on the
distance [29].
2.3.4
Watershed Segmentation
Watershed segmentation technique that starts
from the initial pixel (marker) and then
deluges all the neighboring pixels of that
marker, which known as basins. These basins
ARTÍCULO
2.3.1 Thresholding
Threshold segmentation is one of the simplest
and most common segmentation algorithms
that divide an image based on the gray-scale
value to get the target objects. It can be
classified this methods into two techniques
that include the local threshold and global
threshold techniques such as Otsu's method,
maximum entropy method, and others. The
local threshold technique divides an image
into multiple objects and background by
implementing multiple thresholds. While in the
global threshold technique, an image can be
divided into two target regions and the
background by implementing a single
threshold [23,24]. This technique works with
a threshold value; if the image intensity values
are less than the threshold values, the all
values will be zero (black color). While if the
intensity level values are greater, then the
intensity values will be one (white color).
Thus, these operations will generate the image
in binary mode and will help to get a better
image understanding [25]. Mathematically,
the technique can be written as follows:
514
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
are partition the markers into di๏ฌ€erent regions
using
the
watershed
technique
[30].
Watershed technique used in the medical
image because of the objects (e.g. WBCs) in
the images are overlapping; therefore,
Authors need this kind of algorithms to
separate the overlapping objects and get more
accurate classi๏ฌcation results. Although the
watershed segmentation technique is suitable
for parallel processing, it can produce oversegmentation. Nevertheless, many studies
have been used the marker-controlled
watershed for segmenting the acute leukemia
cells [31]. In marker-controlled watershed
segmentation technique, the boundaries can
be de๏ฌned between external that manually
de๏ฌned according to ROI and internal markers
that automatically obtained. This method will
help to solve the over-segmentation problem
[41]. Mathematically it can be written as
Where the LS (๐‘ฅ) is lower slope of image ๐ผ in ๐‘ฅ
pixel and the (๐‘ฅ) are neighbor pixels. While the
(๐‘ฅ, ๐‘ฆ) is the watershed distance. If ๐‘– = ๐‘—, then
the LS will be zero.
2.3.5
Morphology
Morphology is a technique used to extract
components (ROI) from the medical image for
better description of the regions details [32].
It is deals with the shape of the objects in an
image using specific operations. The basic idea
of morphology technique is move a small
shape that known as the structuring element
over the medical image so that can get the
desired result [33].
The morphological include four operators,
which are the dilation ⊕, erosion โŠ–, opening,
and closing. Can define erosion โŠ– and dilation
⊕ of a binary image as the following:
ARTÍCULO
Where the E is binary image and S is the
structuring element.
515
The opening of an image is erosion followed by
dilation of a binary image as the following:
The closing of an image is dilation followed by
erosion of a binary image as the following:
In
acute
leukemia
detection,
many
researchers have been used the morphological
operation for segmenting the blast cells.
Where it used to enhances the blast cells by
filling gaps and removing the pepper noise
from the cells [34].
2.4
Feature Extraction
Feature extraction (FE) is one of the most
important and critical stages of the image
processing techniques. This stage used to
extract and identify the features derived from
the objects that were segmented from parts of
the image or from the whole image. In other
words, transforming the data that is obtained
from the image into the set of features for
pattern
recognition
is
called
feature
extraction. One of the important issues related
to pattern recognition is choosing the relevant
set of features extraction in order to extract
the relevant information to perform the task
and get accurate information. The features
extraction have been used in many
applications such as leukemia detection,
character recognition, reading bank deposit
slips, applications for credit cards, tax forms,
data entry, check sorting and others [35].
Many features can be extracted from the
objects in the image, such as the shape
features (e.g. area, perimeter, solidity and
others), Texture Features (e.g. homogeneity,
energy, angular second, entropy contrast, and
others), Statistical Features (e.g. mean,
skewness
and,
variance),
Geometrical
features (e.g. perimeter, area, compactnes
and symmetry), color features and so forth
[36].
In acute leukemia detection, the features
extraction stage plays an important role in
determining the leukemia type because blast
cells (ROI) have a lot of information that
included characteristics of nucleus and
cytoplasm [53]. There are different features
have been extracted in the current studies as
shown in Figure 3. The results of feature
extraction stage will be useful for the
classification stage (next stage).
Fig 3: Features extraction techniques.
Feature extraction and feature selection are
crucial steps for image processing and
machine learning techniques. How to extract
and select the ideal features is still a
challenging problem in image processing.
Hence, in this paper, we survey image feature
representation techniques that are used to
extract and select the features from the blast
cell image and then determine whether the
blast cells are cancerous or otherwise.
2.4.1 Color feature
Color feature is one of the most important
relevant information that fetching from the
objects in the image. It is defined based on a
particular model or color space. There are a
number of color spaces such as LUV, HSV, RGB
and others that have been used to make
extracting the features easier. Therefore,
Color features are useful for extraction the
information from the blood cells image for
better classification. [37,38].
2.4.2 Texture feature
Texture feature is a useful characterization for
an image, where pixel properties are used to
measure the color in the image while the
group of pixels is used to measure a texture.
Two techniques are used to extract the texture
features based on the domain, which include
the spectral texture feature extraction and
spatial texture feature extraction. In the
spectral texture feature extraction approach,
an input image is transformed into the
frequency domain, and the texture feature
from the transformed image is then
calculated. While in the spatial approach, the
extracting features have been accomplished
by computing the statistics of a pixel in the
original image domain[39,40]. In the blood
smear images, the texture is an important
feature that used to identify the blast cells by
analyzing that features to get the ROI, which
can help to obtain better classification.
2.4.3 Shape features
Shape feature extraction technique is one of
the key methods in feature extraction field. It
can be classified into two groups, which
include the region-based and contour-based
methods. In the contour-based method, the
shape features are calculating from the
boundary of the shape only, while in the
region-based method, the features are
extracting from the whole region in the image.
Shape features play an important role in acute
leukemia cell detection [41,42].
2.5 Feature Selection
Features selection (FS) is one of most
important technique used to selecting the
more relevant features from the objects in the
image processing. In this case, the algorithms
will select a subset of relevant features, which
will help to increase the accuracy and thus this
process will reduce the cost of computational
[43,44].Therefore,
feature
selections
algorithms will compare all features that are
extracted from the blood smear images and
then selecting the most relevant features.
Different techniques are used for leukemia
cells detection such as PCA (Principal
Component Analysis), GA (Genetic Algorithm),
and PPCA (Probabilistic Principal Component
Analysis) techniques [45].
2.6 Classification
The stage of classification is one of the most
important stages in image processing and
machine learning techniques, and it is an indemand field in this area. Classification is used
to assign and classify a set of unclassified
data. There are two types of classifiers that are
known as supervised and unsupervised
classifications. In the supervised classification,
the set of possible results or classes are known
in advanced. While in the unsupervised
classification, the set of classes are unknown
in advance. Many methods can be used to
form a classification of data that are known as
the classifiers. These classifiers can be used to
classify objects types, such as support vector
machine (SVM), Artificial Neural Network
(ANN), Random forest (RF), KNN (๐พ-Nearest
Neighbor), Naive Bayes (NB), Multilayer
Perceptron (MLP),Hybrid and others, as
summarized in Figure 4 [49]. Once the
features are selected and extracted from the
segmented image, the object's type is
recognized and determined through this stage.
ARTÍCULO
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
516
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
ARTÍCULO
2.6.1 Support vector machine (SVM)
In machine learning and image processing
techniques, the support vector machine (SVM)
also known as the support vector network
[46,47]. It is a supervised learning model that
analyzes the data to use for both classification
and regression tasks. The main objective of
the SVM algorithm is to find a hyper-plane in
N number of features that classify the data
points. The SVM classifier separates the
classes based on the labeled training data. The
main objective of the separate classes is to
find the maximum margin, i.e., find the
maximum distance between the classes. The
maximizing distance provides a capability that
can classify future data points with more
confidence. The SVM classifier is seeking to
find the compromise between the complexities
models according to the training data (limited
sample information). This type of classifier is
suitable for small samples of circumstances.
The SVM classifier algorithm provides four
types of the kernel that include: (1) Sigmoid,
(2) Radial Basis Function, (3) Polynomial and
(4) Linear [50]. The main reason for selecting
SVM for detection of leukemia cell is the
efficiently classify between the normal and
abnormal cells [74].
517
Fig 3: Classification methods for acute
leukemia.
2.6.2 Artificial neural networks
Artificial neural network (ANN) is a type of
artificial intelligence inspired by imitating the
functions from the human brains. In other
words, ANN is an interconnected set of virtual
neurons that created as the computer systems
that work similar to a biological neuron
(electronic probes designed to emulate
neurons). Generally, it consists of a set of
communication nodes that are known as
neurons, which are connected to form a
network of neurons. Each contact between
these neurons has a set of values called
weights that can contribute to determining the
values that are produced by each node of the
network based on the values of this node [48].
The ANN is a framework that has a different
machine learning algorithm that works
together to process complex data. These types
of algorithms depend on the learning to
perform different tasks by using examples that
being programmed for a similar task, i.e.,
working based on the learning. For example,
in the recognition pattern, the systems learn
by analyzing example images to identify other
images such as to identify the image that
contains flowers by analyzing some of the
flowers images that are manually labeled as
"flower" or "no flower”. The systems
automatically
generate
some
of
the
characteristics from the learning material to
identify those objects [49,50].
2.6.3 K-Nearest Neighbor
K-Nearest Neighbor (KNN) is a classifier
technique that widely used, where it uses the
nonparametric to classify data. In this
algorithm, the data are classified by voting
from the nearest neighbors. Based on this
voting, the objects in the image will be
assigned to relevant classes. For acute
leukemia cell classification, the KNN algorithm
is used to get better classification result to
determine if the cells are normal or abnormal
[75].
2.6.4 Random Forest
Random forest (RF) is a classifier that uses an
ensemble learning to classify an object from
the vector. It contains different tree
combinations, which perform voting for the
class, and then selection will be the class who
has a maximum vote [76]. The RF preferred
by the researcher because it corrects the overfitting problem. For acute leukemia detection,
the RF classifier has been found to be suitable
for detection of normal and abnormal cells
[77].
2.6.5 Naive Bayes
Naive Bayes (NB) is a classifier that is used to
classify data based on the features values in
the image that are independent of the other
values [51]; also, the Naive Bayes can be
called as simple Bayes. For acute leukemia
detection, the NB has utilized in current
studies, where it used to classify the WBCs and
determine if the cells are normal or abnormal
[52].
3. Leukemia literature survey
The literature discusses the methods that are
used to segment and detect acute leukemia
using machine learning and image processing
techniques. In this case, the methods detect
acute leukemia using the microscope blood
sample, whereby the process of detection of
acute leukemia cells can be divided into four
stages that include:
๏‚ท
Pre-processing blood image
๏‚ท
Blast cells Segmentation
๏‚ท
Feature Extraction of the blast cell
๏‚ท
Detect and Classify the blast Cell
The steps of pre-processing, segmentation,
feature
extraction,
and
classification
techniques are presented in this paper. All of
these steps are important in detecting
leukemia by using machine learning and
image processing techniques. The success of
each step in the detection of acute leukemia
depends on the previous step.
The work by [23] has proposed several phases
to diagnosis and detect acute lymphoblastic
leukemia from the rest of blood images. This
technique
has
included
blood
image
acquisition using public dataset (ALL-IDB), the
segmentation of the WBCs is accomplished by
a marker-based segmentation, and the
features are extracted from the segmented
WBCs using gray level co-occurrence matrix
(GLCM) with using Probabilistic Principal
Component Analysis (PPCA) to reduce and
obtain the relevant features. This study
proposed the RF classifier to determine if the
cells are normal or abnormal. While the work
by [53] introduced a new approach for
detection of acute leukemia, which includes:
firstly, preprocessing stage by converting RGB
to a gray-scale image and then performs a
histogram equalization to improve and
enhance the image quality to highlights the
contrast
of
nucleus;
secondly,
using
morphological contour segmentation by edges
detection of WBCs, and using HSV conversion
with erosion to segmenting the blast cell;
thirdly, extracting the features of the blast cell
such as geometry, color and texture features;
Finally, classify the acute leukemia cells using
Fuzzy C-means clustering technique.
The work by [54] has proposed the Barebones Particle Swarm Optimization (BBPSO)
technique to diagnosis acute leukemia. This
technique has consists of the following steps:
1) Segmenting the WBCs from the rest of
blood image background using the watershed
technique, 2) Nucleus-cytoplasm separation
phase using a stimulating discriminant
measure
(SDM)-based
clustering,
(3)
Extracting the features such as the statistics,
texture and shape features, 4) Classification
phase using three techniques: 1-Nearest
Neighbour (NN), Radial Basis Function (RBF)
and Support Vector Machine (SVM).
The work by [55] has presented a new
technique used for detection of acute leukemia
using image processing. The proposed
technique is performed by using several
phases. Firstly, the pre-processing by
removing the noise from the images using
median filtering technique followed by unsharp
masking technique and then converting the
blood images from RGB to HSV color space
mode. Secondly, segmenting the RIO using
the thresholding technique (i.e. Otsuโ€Ÿs
method). After that, the features (shape and
texture features) are extracted from the RIO
that has been segmented to classify the cells
whether is cancerous or healthy by the
Support Vector Machine (SVM).
The work by [18] has presented the
segmentation of WBCs using Fuzzy C-Means
clustering for diagnosis acute leukemia. In this
method, the authors are collected the images
from the net, and then the noises are removed
using median filtering technique. In addition,
enhanced the contrast image using the
histogram equalization techniques. In this
work, the blast cell is segmented using Fuzzy
C-means, and then the features (statistical,
texture, and color) are extracted from the ROI
to classify the cells using the SVM.
The work by [57] has proposed the hybrid
hierarchical techniques to detect acute
leukemia. In this technique, the segmentation
of the blast cell by applying the global
thresholding with a Morphological opening on
the image. In the next step, the features (e.g.,
shape and texture) are extracted from the
segmented blast cell, and then these features
are reduced using Principal Component
Analysis (PCA). Finally, the probabilistic neural
network, SVM, K-NN, adaptive neuron-fuzzy
inference system for the classification of the
WBCs on the features that have been
extracted. Table 1 shows the summary and
comparisons
between
the
four
steps
algorithms that are used to detection acute
leukemia.
ARTÍCULO
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
518
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
Table 1. Shows the Summarizes of current methods
Performance
Author,
Year
Methods
-
Preprocessing
J. Su , et al
2017
[7]
Segmentation
This technique used k-means
cluster, and builds cell images to
represent-ing model by HMRF
(Hidden-Markov Random Field)
Feature
Extraction
Classification
-
Preprocessing
Histogram green color of
RGB component that has most
contrast information
Segmentation
Using K-means clustering
technique
S. Negm, et
al 2017
[58]
Geometry features, statistics
features, textures features and
size ratio from regions
selection (Nucleus,
cytoplasm, and whole cell).
Feature
Extraction
Preprocessing
Segmentation
Feature
Extraction
classification
Preprocessing
Segmentation
ARTÍCULO
W.
Srisukkham,
et al 2017
[54]
519
Feature
Extraction
classification
S. Mishra, et
al 2017
[23]
Preprocessing
STRENGTHS:
this method has
improved the
segmentation results in
both of the whole and
cropped images
WEAKNESS:
K values are difficult to
estimate and not suitable
for non-convex clusters.
STRENGTHS:
This technique got good
accuracy by detect ALL
cells and separate
overlapping cells.
Accuracy
This technique achieved a segmentation
accuracy 96% (average) and it is compared
with a manual segmentation that per
implemented by an expert and other existing
methods
The technique achieved the following
accuracies:
The accuracy technique is 99.517%.
The sensitivity accuracy is 99.348%.
WEAKNESS:
K values are difficult to
estimate and it not
suitable for non-convex
clusters.
The specificity accuracy is 99.529%.
Artificial Neural Network
(ANN)
classification
J. Rawat. , et
al,
2017
[8]
STRENGTHS
And
WEAKNESS
Convert RGB image into gray scale
image. Then enhance the blood
image quality using histogram
equalization
STRENGTHS:
This method used GAGaussian with radial
basis kernel (RBF) for
classify acute leukemia
Divide the image to sub-images and
used Thresholding using Otsu’s
technique
WEAKNESS:
- need high contrast
between the objects and
background
- It is very sensitive to
gray-scale unevenness
and noise in the image
This method used the statistical,
texture , and geometrical features
This method got accuracy for the classification
of leukemia cells 99.5% .
The accuracy for the FAB subtypes of ALL is
97.1%.
The accuracy for the FAB subtypes of AML is
98.5%
genetic algorithm GA-Gaussian
radial basis kernel
converted the image to Sub-images
Using marker-controlled watershed
to getting the blast cells
Particle Swarm Optimization
(PSO), Dragonfly Algorithm
(DA)and Cuckoo Search (CS)
STRENGTHS:
this method has good
performance by using
marker-controlled
watershed with Particle
Swarm Optimization
SVM
WEAKNESS:
- over segmentation
- sensitivity to noise
using histogram equalization and
weiner filtering to enhance blood
images
STRENGTHS:
marker-based watershed
segmentation algorithm
This technique achieved
the performances of 94.94% and 96.25% as
accuracy
The presented method achieved results as
following :
The segmentation accuracy , 96.29%
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
Segmentation
By using marker-based watershed
segmentation scheme to separate
the overlapping cells
Feature
Extraction
gray level co-occurrence matrix
(GLCM)
and the probabilistic principal
component analysis (PPCA) is used
to extract the relevant Feature
reduction
Preprocessing
T.
Karthikeyan,
et al 2017
[18]
Feature
Extraction
Gabor Texture Extraction method is
used to extract color
Features
Segmentation
A. Gajul, et
al
2016
[56]
Feature
Extraction
classification
Preprocessing
Segmentation
M. Sukanya,
et al 2016
[60]
Feature
Extraction
classification
K.
ElDahshan,
et al 2015
[59]
-Image acquisition
-Noise removal using Adaptive
median filter
-Contrast enhancement using
histogram equalization
Fuzzy c- means clustering
Preprocessing
Preprocessing
Segmentation
WEAKNESS:
- over segmentation
- sensitivity to noise
Random Forest (RF) technique
Segmentation
classification
The classification accuracy is 99.004%
STRENGTHS:
this technique got high
accuracy by used Fuzzy
c-mean with SVM
WEAKNESS:
- determination of fuzzy
membership is complex.
- Sensitive to the noises
in images.
- Computationally
expensive due to long
time.
The Fuzzy c-means method gives 90% as
accuracy
Support Vector Machine (SVM)
Convert RGB image to CIELAB
L*a*b* color space.
k-means clustering
-Hausdorff Dimension (HD
- Local Binary Pattern (LBP)
-Shape Features:
-GLCM Features:
-Color Feature: Cell Energy
STRENGTHS:
This good technique to
detect AML and separate
the overlapping cell by
using K-mean clustering
with SVM
WEAKNESSES:
- K values are difficult to
estimate - time
consuming
- it is not suitable for
non-convex sclusters.
This technique achieved result as following :
-Sensitivity 71.43 %
-Specificity 100 %
-Precision 100 %
-overall classification performance of 83.33 %
support vector machines (SVM)
Convert RGB image to CIEL*a*b
color space
K-Means Clustering.
This technique used the edgetexture feature, Discriminative
Robust Local Binary Pattern
(DRLBP) and Discriminative
Robust Local ternary Pattern
(DRLTP)
STRENGTHS This
technique used K-Means
Clustering with DRLBP
and DRLTP to gives a
perfect decision about
the disease.
This technique presented system performs well
WEAKNESS:
- K values are difficult
to estimate - time
consuming
- it is not suitable for
non-convex sclusters.
support vector machines (SVM)
Convert RGB to HSV color space
STRENGTHS:
This technique used the
FPGA with Xilinx for
segmentation
-
ARTÍCULO
classification
is used to segmentation
and separation
overlapping cells and it
is achieved good
accuracy
520
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
-FPGA pre-processing
-Xilinx models for FPGA
segmentation
-FPGA post-processing
Feature
Extraction
-
classification
Preprocessing
G. Patil, et al
2015
[55]
Segmentation
Feature
Extraction
Shape features and
Densitom-etric features: Energy,
GLCM, Correlation
Preprocessing
.T.
Madhloom,
et al 2012
[61]
convert the image from RGB color
space to HSV domain and perform
the median filtering
This method have used the
Thresholding is done by Otsu’s
method with bounding box
technique to getting sub-image
classification
Segmentation
This technique used CIELAB Color
Space, Transform to HSV Color
Space and extract (H) Channel
Marker- Controlled Watershed
and SRG algorithm
-
classification
Segmentation
L. Putzu, et
al 2014
[21]
Feature
Extraction
ARTÍCULO
classification
521
M.
MoradiAmin
, et al 2015
[12]
Preprocessing
Segmentation
STRENGTHS:
Its fast ,simple and got
good result
WEAKNESS:
- need high contrast
between the objects and
background
-it is very sensitive to
gray-scale unevenness
and noise in the image
Support Vector Machine(SVM)
Feature
Extraction
Preprocessing
WEAKNESS:
- over segmentation
- sensitivity to noise
Convert the images from RGB to
CMYK colour model, then perform
the Histogram equalization to
improve the image quality
threshold with Zack algorithm,
and then separation of overlapping
cells by
watershed segmentation technique
with Solidity to remove the nois
-shape features
-color features
-texture features
STRENGTHS
this method give good
accuracy for
segmentation by using
Marker-Controlled
Watershed and
SRG algorithm
This technique presented segmentation results
of 96% accuracy
WEAKNESS:
- over segmentation
- sensitivity to noise
STRENGTHS: this
technique achieved good
accuracy by using the
Zack algorithm with
SVM .
Using this technique achieved the results as
following :
WEAKNESS:
- over segmentation
- sensitivity to noise
- need high contrast
image
as accuracy from 33images is 92%
identification of ALL, with an accuracy 93%
and sensitivity 98%
support vector machine (SVM)
Convert RGB to HSV Color space,
and then Histogram equalization on
V band
By using fuzzy c-means clustering
and then using watershed algorithm
to separate overlapping objects
STRENGTHS:
This technique achieved
a good accuracy by used
fuzzy c-mean with SVM
.Also its success to find
the sub-type of ALL.
The presented method achieved results as
following :
The sensitivity , 98%
The specificity
95 %
The classification accuracy is 97%
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
classification
support vector machine (SVM)
4. Analysis and Discussion
A survey of image processing techniques that
are used for segmentation and classification of
acute leukemia cells have presented in this
paper. Moreover, we have discussed the
techniques and methods that have used in
previous works. To check how the previous
works were implemented and compared these
methods to the manual methods for leukemia
cells detection. These studies focused on the
results of each following steps: preprocessing,
segmenting, features extracting, and then the
detection
of
acute
leukemia.
The
computerized methods are more reliable,
accurate, and efficient compared to the
manual techniques that were time-consuming
and less efficient.
In the segmentation stage, many techniques
have been proposed by previous studies to
obtain the ideal results for segmenting and
detecting of acute leukemia cells. The higher
results of this phase were achieved using a
marker-based
watershed
technique
to
segment and separate the overlapping cells
with 96.29% of accuracy. In addition, the
researchers
also
use
some
of
the
segmentation techniques such as Fuzzy CMeans and K-Means, thresholding, watershed,
HSV color based and others to segment the
blast cell accurately. Most of these methods
have difficulties in segmenting complex blood
images due to some major shortcomings, such
as lack of contrast between objects and
background and sensitivity to noise in images.
Therefore, it is indispensable to develop an
effective technique for segmenting blast cells
from peripheral blood smears. In addition, the
scientists are still seeking for a more accurate
technique to segment the blast cells.
In the phase of feature extraction, various
features used include the color, geometrical
and texture features. The numbers of the
features are numerous; therefore, it is better
to make the feature selection by using T-SNE,
LDA or PCA. Fisher's Discrimination Ratio
(FDR) can also be used followed by Exhaustive
Search (ES) to obtain the best extraction
features.
WEAKNESS:
- determination of fuzzy
membership is complex.
- Sensitive to the noises
in images.
- Computationally
expensive due to long
time.
The work by W. Srisukkham, et al [54] have
divided the image to Sub-images for
preprocessing of blood smear images and then
marker-controlled watershed was used to
detect and segment the blast cells and after
extracting features using Particle Swarm
Optimization (PSO), Dragonfly Algorithm (DA)
and Cuckoo Search (CS); the SVM was used
for detection and classification of blasted cells.
This method able to achieve 94.94% as
accuracy result. While the work by S. Negm,
et al [58] have improved the accuracy to
99.517% using Histogram green color of RGB
component for preprocessing of blood smear
images and then Using K-means clustering
technique was used to segment the blast cells
and after extracting features using Geometry
features, statistics features, and textures
features; the Artificial Neural Network (ANN)
was used for detection and classification of
blasted cells.
In
the
classification
phase,
different
techniques were carried out by researchers to
obtain the accurate results, where the ANN
and SVM techniques were the most accurate.
Moreover, the detection phase was dependent
on the previous stages (segmentation and
feature extraction). The studies that used
Support
Vector
Machine
(SVM)
have
presented higher accuracy compared to other
classifiers. Moreover, other studies also found
the Fuzzy logic-based classifier was greater
than SVM in terms of accuracy. Therefore, it is
indispensable to develop an effective classifier
to detect blast cells from peripheral blood
smears images.
After we have analyzed the previous
techniques, we have come to the main point:
the scientists still need to seek for a more
accurate technique to detect acute leukemia
cells. Acute leukemia detection is a very
sensitive issue and it is related to humans
health. Therefore, the Accuracy of diagnosis
process should be flawless in order to replace
human
operators
with
computerized
diagnoses. But the replacement is a very
challenging task because the complexities of
blood cells are high.
5. Limitations
ARTÍCULO
Feature
Extraction
- geometric feature for size and
shape of a nucleus
- statistical features for gray scale
image histogram
522
ARTÍCULO
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
523
Although the database of sources that we have
used are reliable; but the identification was
difficult. Moreover, the timeliness of this study
was one of limitations because of the
increasing progress in this area. In addition,
the response of the researchers to this field is
the objective of this review.
6. Conclusion
Many studies have been presented in this
paper. These studies were conducted for the
segmentation and detection of acute leukemia
by using image processing and machine
learning
techniques.
Image
processing
techniques can be easily and rapidly diagnose
leukemia, and this increases the chance of
patients to be saved and allowing them to be
given the appropriate treatment. Image
processing and machine learning techniques
may replace blood analysis experts in the
detection and classification of leukemia in
terms of accuracy and time.
In future work, researchers could focus on
other methods of segmenting acute leukemia
cells such as using a combination of two or
three segmentation techniques to make full
use of the different algorithms advantages to
achieve better segmentation for the blast
cells. In addition, we can suggest using a
machine learning algorithm (e.g., support
vector machine, neural network, and others)
to improve the segmentation of the blast cell.
Moreover, there will be more medical image
segmentation techniques widely used; as well
as more than one chance to find other suitable
features that can detect the type of Acute
Leukemia. The features selection techniques
and dimensional reduction can be performed
to achieve greater efficiency.
6. Acknowledgment
We would like to thank the Higher Education
Minatory
at
Malaysia
and
University
Technology Malaysia for their educational and
financial support. This work is conducted
under Cyber Physical Systems Research Group
(CPSRG) at Razak Faculty of Technology and
Informatics that funded by University
Technology Malaysia (UTM).
References
[1]. A. A. El-Nasser, M. Shaheen, and H. ElDeeb, “Enhanced leukemia cancer
classifier algorithm,” in Science and
Information Conference (SAI), 2014,
2014, pp. 422–429..
[2]. J. E. Sadler, “What’s new in the diagnosis
and pathophysiology of thrombotic
thrombocytopenic purpura,” ASH Educ.
[3].
[4].
[5].
[6].
[7].
[8].
[9].
[10].
[11].
[12].
Program Book, vol. 2015, no. 1, pp.
631–636, 2015.
G. Singh, G. Bathla, and S. Kaur, “Design
of New Architecture to detect leukemia
cancer from medical images,” Int J Appl
Eng Res, vol. 11, no. 10, pp. 7087–7094,
2016.
J. Rawat, A. Singh, H. S. Bhadauria, and
J. Virmani, “Computer aided diagnostic
system for detection of leukemia using
microscopic images,” Procedia Comput.
Sci., vol. 70, pp. 748–756, 2015.
S. S. Al-jaboriy, N. N. A. Sjarif, S.
Chuprat, and W. M. Abduallah, “Acute
Lymphoblastic Leukemia Segmentation
Using Local Pixel Information,” Pattern
Recognit. Lett., 2019.
G. Biji and D. S. Hariharan, “White Blood
Cell
Segmentation
Techniques
in
Microscopic
Images
for
Leukemia
Detection,” IOSR J. Dent. Med. Sci., vol.
15, no. 2, pp. 45–51, 2016.
Su, J., Liu, S., & Song, J. (2017). A
segmentation method based on HMRF
for the aided diagnosis of acute myeloid
leukemia. Computer Methods
and
Programs in Biomedicine, 152, 115–123
Rawat, J., Singh, A., Bhadauria, H. S.,
Virmani, J., & Devgun, J. S. (2017).
Computer
assisted
classification
framework for prediction of acute
lymphoblastic and acute myeloblastic
leukemia. Biocybernetics and Biomedical
Engineering, 37(4), 637–654.
Shaikh, M. B. N., & Deshpande, S.
(2017). Computer aided leukemia
detection using digital image processing
techniques. In Recent Trends in
Electronics,
Information
&
Communication Technology (RTEICT),
2017 2nd IEEE International Conference
on (pp. 344–348). IEEE
Yogamangalam, R., & Karthikeyan, B.
(2013).
Segmentation
techniques
comparison
in
image
processing.
International Journal of Engineering and
Technology (IJET), 5(1), 307–313.
Akilandeswari, U., Nithya, R., & Santhi,
B. (2012). Review on feature extraction
methods
in
pattern
classification.
European Journal of Scientific Research,
71(2), 265–272.
MoradiAmin, M., Samadzadehaghdam,
N., Kermani, S., & Talebi, A. (2015).
Enhanced
recognition
of
acute
lymphoblastic
leukemia
cells
in
[13].
[14].
[15].
[16].
[17].
[18].
[19].
[20].
[21].
microscopic images based on feature
reduction using principle component
analysis.
Frontiers
in
Biomedical
Technologies, 2(3), 128–136
Nasir, A. A., Mashor, M. Y., & Hassan, R.
(2012). Leukaemia screening based on
fuzzy ARTMAP and simplified fuzzy
ARTMAP neural networks. In Biomedical
Engineering and Sciences (IECBES),
2012 IEEE EMBS Conference on (pp. 11–
16). IEEE.
Rawat, J., Bhadauria, H. S., Singh, A., &
Virmani, J. (2015). Review of leukocyte
classification techniques for microscopic
blood images. In Computing for
Sustainable
Global
Development
(INDIACom), 2015 2nd International
Conference on (pp. 1948–1954). IEEE.
Neoh, S. C., Srisukkham, W., Zhang, L.,
Todryk, S., Greystoke, B., Lim, C. P., …
Aslam, N. (2015). An intelligent decision
support system for leukaemia diagnosis
using
microscopic
blood
images.
Scientific Reports, 5.
Goutam, D., & Sailaja, S. (2015).
Classification of acute myelogenous
leukemia in blood microscopic images
using
supervised
classifier.
In
Engineering and Technology (ICETECH),
2015 IEEE International Conference on
(pp. 1–5). IEEE
Amin, M. M., Kermani, S., Talebi, A., &
Oghli, M. G. (2015). Recognition of acute
lymphoblastic
leukemia
cells
in
microscopic images using k-means
clustering and support vector machine
classifier. Journal of Medical Signals and
Sensors, 5(1), 49.
Karthikeyan, T., & Poornima, N. (2017).
Microscopic Image Segmentation Using
Fuzzy C Means For Leukemia Diagnosis.
Leukemia, 4(1).
Viswanathan, P. (2015). Fuzzy C means
detection
of
leukemia
based
on
morphological contour segmentation.
Procedia Computer Science, 58, 84–90.
Francis, E. U., Mashor, M. Y., Hassan, R.,
& Abdullah, A. A. (2011). Screening of
bone marrow slide images for leukemia
using multilayer perceptron (MLP). In
Industrial Electronics and Applications
(ISIEA), 2011 IEEE Symposium on (pp.
643–648). IEEE.
Putzu, L., Caocci, G., & Di Ruberto, C.
(2014). Leucocyte classification for
leukaemia
detection
using
image
[22].
[23].
[24].
[25].
[26].
[27].
[28].
[29].
[30].
processing
techniques.
Artificial
Intelligence in Medicine, 62(3), 179–
191.
Alférez Baquero, E. S., Merino, A., Mújica
Delgado, L. E., Ruiz Ordóñez, M.,
Bigorra, L., & Rodellar Benedé, J.
(2013). Digital blood image processing
and fuzzy clustering for detection and
classification of atypical lymphoid B cells.
In Jornades de recerca EUETIB (pp. 1–
12). EUETIB.
Mishra, S., Majhi, B., Sa, P. K., &
Sharma, L. (2017). Gray level cooccurrence matrix and random forest
based acute lymphoblastic leukemia
detection. Biomedical Signal Processing
and Control, 33, 272–280.
L. S. Davis, A. Rosenfeld, and J. S.
Weszka, “Region extraction by averaging
and thresholding,” IEEE Trans. Syst. Man
Cybern., no. 3, pp. 383–388, 1975.
N. R. Pal and S. K. Pal, “A review on
image
segmentation
techniques,”
Pattern Recognit., vol. 26, no. 9, pp.
1277–1294, 1993.
M. A. Wani and B. G. Batchelor, “Edgeregion-based segmentation of range
images,” IEEE Trans. Pattern Anal. Mach.
Intell., vol. 16, no. 3, pp. 314–319,
1994.
Ayman Dawood Salman1, Osamah
Ibrahim Khalaf and Ghaida Muttashar
Abdulsahib,
2019.
An
adaptive
intelligent alarm system for wireless
sensor network. Indonesian Journal of
Electrical Engineering and Computer
Science, Vol. 15, No. 1, July 2019, pp.
142~147
K.-S. Chuang, H.-L. Tzeng, S. Chen, J.
Wu, and T.-J. Chen, “Fuzzy c-means
clustering with spatial information for
image segmentation,” Comput. Med.
Imaging Graph., vol. 30, no. 1, pp. 9–
15, 2006.
M. E. Celebi, H. A. Kingravi, and P. A.
Vela, “A comparative study of efficient
initialization methods for the k-means
clustering algorithm,” Expert Syst. Appl.,
vol. 40, no. 1, pp. 200–210, 2013.
N. Senthilkumaran and R. Rajesh,
“Image segmentation-a survey of soft
computing
approaches,”
in
2009
International Conference on Advances in
Recent Technologies in Communication
and Computing, 2009, pp. 844–846.
ARTÍCULO
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
524
ARTÍCULO
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
525
[31]. M. K. Kundu and S. K. Pal, “Thresholding
for edge detection using human
psychovisual
phenomena,”
Pattern
Recognit. Lett., vol. 4, no. 6, pp. 433–
441, 1986
[32]. J. C. Bezdek, L. O. Hall, and L. Clarke,
“Review of MR image segmentation
techniques using pattern recognition,”
Med. Phys., vol. 20, no. 4, pp. 1033–
1048, 1993.
[33]. L. Weisz, “Pattern Recognition Statistical
Structural And Neural Approaches,”
Pattern Recogn, vol. 1, no. 2, 2016
[34]. D. L. Pham, C. Xu, and J. L. Prince,
“Current methods in medical image
segmentation,” Annu. Rev. Biomed.
Eng., vol. 2, no. 1, pp. 315–337, 2000.
[35]. Mohammed,
M.A.,
Ghani,
M.K.A.,
Arunkumar, N.A., Hamed, R.I., Mostafa,
S.A., Abdullah, M.K. and Burhanuddin,
M.A., 2018. Decision support system for
nasopharyngeal
carcinoma
discrimination from endoscopic images
using artificial neural network. The
Journal
of
Supercomputing,
https://doi.org/10.1007/s11227-0182587-z.
[36]. H. Motoda and H. Liu, “Feature selection,
extraction and construction,” Commun.
IICM Inst. Inf. Comput. Mach. Taiwan
Vol, vol. 5, pp. 67–72, 2002.
[37]. L. Ladha and T. Deepa, “Feature
selection methods and algorithms,” Int.
J. Comput. Sci. Eng., vol. 3, no. 5, pp.
1787–1797, 2011.
[38]. D. ping Tian, “A review on image feature
extraction
and
representation
techniques,” Int. J. Multimed. Ubiquitous
Eng., vol. 8, no. 4, 2013
[39]. A. K. Jain and A. Vailaya, “Image
retrieval using colour and shape”,
Pattern Recognition, vol. 29, no. 8,
(1996), pp. 1233-1244.
[40]. M. Flickner, H. Sawhney, W. Niblack, et
al., “Query by image and video content:
the QBIC system”, IEEE Computer, vol.
28, no. 9, (1995), pp. 23-32.
[41]. Osamah Ibrahim Khalaf, Bayan Mahdi
Sabbar''An overview on wireless sensor
networks and finding optimal location of
node'',Periodicals of Engineering and
Natural Sciences, Vol 7, No 3 (2019)
[42].
[43]. J. Huang, S. Kuamr, M. Mitra, et al.,
“Image
indexing
using
colour
[44].
[45].
[46].
[47].
[48].
[49].
[50].
[51].
[52].
correlogram”, In Proc. CVPR, (1997), pp.
762-765.
Osamah
Ibrahim
Khalaf,
Ghaida
Muttashar Abdulsahib and Muayed
Sadik, 2018. A Modified Algorithm for
Improving Lifetime WSN. Journal of
Engineering and Applied Sciences, 13:
9277-9282
D. Zhang and G. Lu, “Review of shape
representation
and
description
techniques”, Pattern Recognition, vol.
37, no. 1, (2004), pp. 1-19.
Mohammed,
M.A.,
Ghani,
M.K.A.,
Arunkumar, N.A., Hamed, R.I., Abdullah,
M.K. and Burhanuddin, M.A., 2018. A
real time computer aided object
detection of nasopharyngeal carcinoma
using genetic algorithm and artificial
neural network based on Haar feature
fear. Future Generation Computer
Systems, 89, pp.539-547.
Veerabhadrappa, L. Rangarajan, “Bilevel Dimensionality Reduction Methods
Using Feature Selection and Feature
Extraction”, International Journal of
Computer Applications, vol. 4(2), pp. 3338, 2010.
L. Yu, and H. Liu, “Feature Selection for
High-Dimensional
Data:
A
Fast
Correlation-Based Filter Solution”, In:
Proceeding of th
S. Khalid, T. Khalil, and S. Nasreen, “A
survey of feature selection and feature
extraction
techniques
in
machine
learning,” in Science and Information
Conference (SAI), 2014, 2014, pp. 372–
378.
M. A. Alsalem et al., “A Review of the
Automated Detection and Classification
of
Acute
Leukaemia:
Coherent
Taxonomy, Datasets, Validation and
Performance Measurements, Motivation,
Open
Challenges
and
Recommendations,” Comput. Methods
Programs Biomed., 2018.
Ghani, M.K.A., Mohamed, M.A., Mostafa,
S.A., Mustapha, A., Aman, H. and Jaber,
M.M., 2018. The Design of Flexible
Telemedicine Framework for Healthcare
Big Data. International Journal of
Engineering & Technology, 7(3.20),
pp.461-468.
S.Arunadevi1dr. S. Daniel Madan Raja “A
Survey
On
Image
Classification
Algorithm
Based
On
Per-Pixel”
International Journal Of Engineering
[53].
[54].
[55].
[56].
[57].
[58].
[59].
[60].
[61].
Research And General Science Vol 2,
Issue 6, October-November, 2014 .
D. Lu and Q. Weng, “A survey of image
classification methods and techniques
for
improving
classification
performance,” Int. J. Remote Sens., vol.
28, no. 5, pp. 823–870, 2007
Mohammed,
M.A.,
Ghani,
M.K.A.,
Hamed, R.I. and Ibrahim, D.A., 2017.
Review on Nasopharyngeal Carcinoma:
Concepts,
methods
of
analysis,
segmentation, classification, prediction
and impact: A review of the research
literature. Journal of computational
science, 21, pp.283-298.
W. Srisukkham, L. Zhang, S. C. Neoh, S.
Todryk, and C. P. Lim, “Intelligent
leukaemia diagnosis with bare-bones
PSO based feature optimization,” Appl.
Soft Comput., vol. 56, pp. 405–419,
2017.
T. G. Patil and V. B. Raskar, “Automated
Leukemia Detection By Using Contour
Signature Method,” Int. J. Adv. Found.
Res. Comput., vol. 2, no. 6, 2015.
Y. A. Gajul and R. Shelke, “Computerized
Detection System for Acute Myelogenous
Leukemia in Blood Microscopic Images,”
Int. J. Innov. Res. Sci. Eng. Technol.
June, 2016.
Rawat, J., Singh, A., Bhadauria, H. S.,
Virmani, J., & Devgun, J. S. (2017).
Leukocyte Classification using Adaptive
Neuro-Fuzzy
Inference
System
in
Microscopic Blood Images. Arabian
Journal for
Arunkumar, N., Mohammed, M.A.,
Ghani, M.K.A., Ibrahim, D.A., Abdulhay,
E., Ramirez-Gonzalez, G. and de
Albuquerque, V.H.C., 2018. K-Means
clustering and neural network for object
detecting and identifying abnormality of
brain
tumor.
Soft
Computinghttps://doi.org/10.1007/s00
500-018-3618-7.
K. ElDahshan, M. Youssef, E. Masameer,
and M. A. Mustafa, “An efficient
implementation of acute lymphoblastic
leukemia images segmentation on the
FPGA,” Adv. Image Video Process., vol.
3, no. 3, p. 8, 2015.
Mostafa, S.A., Mustapha, A., Khaleefah,
S.H., Ahmad, M.S. and Mohammed,
M.A., 2018, February. Evaluating the
performance of three classification
methods in diagnosis of Parkinson’s
[62].
[63].
[64].
[65].
[66].
[67].
[68].
disease. In International Conference on
Soft Computing and Data Mining (pp. 4352). Springer, Cham.
H. T. Madhloom, S. A. Kareem, and H.
Ariffin, “A robust feature extraction and
selection method for the recognition of
lymphocytes versus acute lymphoblastic
leukemia,” in Advanced Computer
Science Applications and Technologies
(ACSAT), 2012 International Conference
on, 2012, pp. 330–335.
A. R. Begum and T. A. Razak, “A
Proposed Novel Method for Detection
and Classification of Leukemia using
Blood Microscopic Images.,” Int. J. Adv.
Res. Comput. Sci., vol. 8, no. 3, 2017
S. Agaian, M. Madhukar, and A. T.
Chronopoulos, “Automated screening
system for acute myelogenous leukemia
detection in blood microscopic images,”
IEEE Syst. J., vol. 8, no. 3, pp. 995–
1004, 2014.
Ghani,
M.K.A.,
Mohammed,
M.A.,
Arunkumar, N., Mostafa, S.A., Ibrahim,
D.A., Abdullah, M.K., Jaber, M.M.,
Abdulhay, E., Ramirez-Gonzalez, G. and
Burhanuddin, M.A., 2019. Decision-level
fusion scheme for nasopharyngeal
carcinoma identification using machine
learning techniques. Neural Computing
and
Applications,
https://doi.org/10.1007/s00521-0183882-6.
R. Bhattacharjee and L. M. Saini,
“Robust technique for the detection of
acute lymphoblastic leukemia,” in
Power, Communication and Information
Technology Conference (PCITC), 2015
IEEE, 2015, pp. 657–662.
Arunkumar, N., Mohammed, M.A.,
Mostafa, S.A., Ibrahim, D.A., Rodrigues,
J.J. and de Albuquerque, V.H.C., 2018.
Fully
automatic
modelโ€based
segmentation
and
classification
approach for MRI brain tumor using
artificial neural networks. Concurrency
and
Computation:
Practice
and
Experience,
https://doi.org/10.1002/cpe.4962.
Mostafa,
S.A.,
Mustapha,
A.,
Mohammed,
M.A.,
Hamed,
R.I.,
Arunkumar, N., Ghani, M.K.A., Jaber,
M.M. and Khaleefah, S.H., 2019.
Examining multiple feature evaluation
and classification methods for improving
the diagnosis of Parkinson’s disease.
ARTÍCULO
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
526
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
[69].
[70].
[71].
[72].
ARTÍCULO
[73].
527
Cognitive Systems Research, 54, pp.9099.
R. J. A. Cabrera, C. A. P. Legaspi, E. J.
G. Papa, R. D. Samonte, and D. D. Acula,
“HeMatic: An automated leukemia
detector with separation of overlapping
blood cells through Image Processing
and Genetic Algorithm,” in Applied
System
Innovation
(ICASI),
2017
International Conference on, 2017, pp.
985–987.
M. A. Khosrosereshki and M. B. Menhaj,
“A fuzzy based classifier for diagnosis of
acute lymphoblastic leukemia using
blood smear image processing,” in Fuzzy
and Intelligent Systems (CFIS), 2017
5th Iranian Joint Congress on, 2017, pp.
13–18.
Abdulhay,
E.,
Mohammed,
M.A.,
Ibrahim, D.A., Arunkumar, N. and
Venkatraman, V., 2018. Computer aided
solution for automatic segmenting and
measurements of blood leucocytes using
static microscope images. Journal of
medical systems, 42(4), p.58.
Mohammed,
M.A.,
Ghani,
M.K.A.,
Arunkumar, N., Mostafa, S.A., Abdullah,
M.K. and Burhanuddin, M.A., 2018.
Trainable model for segmenting and
identifying Nasopharyngeal carcinoma.
Computers & Electrical Engineering, 71,
pp.372-387.
Mohammed, M.A., Al-Khateeb, B.,
Rashid, A.N., Ibrahim, D.A., Ghani,
M.K.A. and Mostafa, S.A., 2018. Neural
network and multi-fractal dimension
features for breast cancer classification
[74].
[75].
[76].
[77].
[78].
from ultrasound images. Computers &
Electrical Engineering. 70, pp. 871-882.
Ibrahim Obaid, O., Abed Mohammed, M.,
Khanapi Abd Ghani, M., A. Mostafa, S., &
Taha AL-Dhief, F. (2018). Evaluating the
Performance
of
Machine
Learning
Techniques in the Classification of
Wisconsin Breast Cancer. International
Journal of Engineering & Technology,
7(4.36), 160-166.
J. Laosai and K. Chamnongthai, “Acute
leukemia classification by using SVM and
K-Means clustering,” in Proceedings of
the
2014
International
Electrical
Engineering Congress, iEECON 2014,
Thailand, March 2014. [75]
Ogudo, K.A.; Muwawa Jean Nestor, D.;
Ibrahim Khalaf, O.; Daei Kasmaei, H. A
Device Performance and Data Analytics
Concept for Smartphones’ IoT Services
and Machine-Type Communication in
Cellular Networks. Symmetry 2019, 11,
593.
Mohammed,
M.A.,
Ghani,
M.K.A.,
Hamed, R.I. and Ibrahim, D.A., 2017.
Analysis of an electronic methods for
nasopharyngeal carcinoma: Prevalence,
diagnosis, challenges and technologies.
Journal of computational science, 21,
pp.241-254.
M. Saraswat and K. V. Arya, “Feature
selection and classification of leukocytes
using random forest,” Medical &
Biological Engineering & Computing, vol.
52, no. 12, pp. 1041–1052, 2014.
Table 1: Examples State of Tablet's Pill
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
Max
Mean
Mean
Mean
Mean
Mean
Mean
Mean
Mean
Mean
Mean
ge
Pill 1
Pill 2
Pill 3
Pill 4
pill5
Pill 6
Pill 7
Pill 8
Pill 9
Pill
pill
.
10
*75
pi
.No
mean
State
Ima
no
%
1
97.11
99.70
92.23
98.87
92.83
96.01
100.0
4
1
8
5
5
9
0
99.62
99.90
100.0
98.50
96.11
99.40
94.13
98.76
0
0
7
9
2
3
2
99.77
100.0
97.75
5
0
98.00
99.40
99.10
88.25
99.10
89.75
99.50
97.31
99.88
99.70
2
4
8
4
12
7
34
75
1
99.71
99.60
98.31
99.30
98.80
98.70
100.0
99.40
98.60
2
1
2
3
5
64
0
29
69
99.88
99.32
98.20
98.50
99.00
99.20
98.60
94.31
98.87
7
2
0
7
4
39
6
1
51
100.0
98.98
99.77
99.77
100.0
98.40
98.60
100.0
99.80
0
7
5
5
0
79
9
0
0
98.61
97.63
99.11
48.35
99.43
98.31
97.91
98.27
99.11
74.57
2
7
7
0
75
2
04
8
7
8
88.44
95.72
92.46
42.29
90.43
92.93
94.93
86.95
90.88
96.06
72.04
4
5
3
4
8
53
8
16
8
2
7
2
2
3
Safety
99.32
98.76
9
0
75.00
75.00
1
0
10
Safety
10
Safety
74.91
Dama
4
75.00
9
74.91
9
ge
Dama
5
6
7
8
5
75.00
Dama
9
ge
Dama
1
0
ge
Dama
1
0
ge
Experimental Results
In this section, the performance of the
implemented system is tested for safety and
damage tablet images. The first experiment is
tested on fitted tablet images. The second
experiment is tested on damage tablet images
include broken one of the pills. The third
experiment is tested on damage tablet images
include missing one of the pills from the tablet.
All the experiments are performed Matlab in
32-bit system with 2.40 GHz core i3 processor
and 3 GB of RAM, run with the MS Win.7
operating system.
The pill image inputted into inspection green
circle pill system and pass through all stages
as described in the following stages:
Stage 1:
Input pill image mentioned above as
shown in Figure 4, to pass through
preprocessing steps.
Figure 4: Original green tablet images
Stage2:
The results of preprocessing step shown in
Figure 5.
ARTÍCULO
5.
97.07
ge
5
528
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
descriptors lead to correct circle detection. The
obtained results indicated that this method
leading to high speed, high precision, noise
resistance and false fault detection. This
system is time efficient as it requires just
0.975 seconds on an average for executing the
program and predicting the result.
The performance of an inspection green pill
system is 100% for each database that
contain a) 50 image of the tablet fill pill and
not broken. b) 50 image of the tablet empty
one of the pills. c) 50 image of the tablet
broken one of the pills. Noticeable the
performance of an inspection green bell
system Excellent and nearer from Human
vision (The person checking the manual).
References
[1]
Figure 5: preprocessing green tablet images
Stage3: The results of Pill Radii for circle
detection step shown in Table 2.
Table 2: PILL RADII
No.
[2]
1
2
3
4
5
6
7
8
9
10
18
18
18
16
18
18
18
17
17
19
[3]
Pill
radii
Stage4: analysis circle detection of green pill,
table 3 shows the mean of pill detection.
[4]
Table 3: Mean of Pill Detection
no.pill
1
2
3
4
5
6
7
8
9
10
mean
99.9
100
99.4
96.9
98.2
98.6
99.0
99.00
98.4
99.5
pill
Stage5:
Print number of detected circles & print
number of broken circles shown in figure 6
[5]
[6]
ARTÍCULO
Figure 6: Eventual Image of the proposed
system
6. Conclusion
529
The experimental results indicated that CHT
can be effectively applied to detect the centers
and radii of pill circle and color transforms that
led to higher enhance the contrast and
emphasis of green color for image these
[7]
Deepti, R. B. (2015). Enhanced Feature
Extraction Technique for Detection of
Pharmaceutical Drugs. Volume 3, Issue
3.
Dhiman, M. K., & Gupta, D. R. (2014).
Detection of Broken Blister using Canny
and Rc-algorithm International Journal
of Scientific Research Engineering &
Technology (IJSRET), Vol. 3, Issue 3.
Jiang, Y., Ma, S., Gao, H., & Xia, F.
(2014). Research on defect detection
technology of tablets in aluminum plastic
package. The Open Automation and
Control Systems Journal, 6, 940-951.
Abha
Sharma1,
Sugandha
Arora.
(2012).Inspection and Classification of
Defects in Pharmaceutical Capsules
Using Neural Network, ISSN: 2278067X, Volume 1, Issue 10.
Ramya, S., Suchitra, J., & Nadesh, R. K.
(2013).
Detection
of
broken
pharmaceutical drugs using enhanced
feature
extraction
technique.
International Journal of Engineering and
Technology, 5 (2), 1407-1411.
Ayman Dawood Salman1, Osamah
Ibrahim Khalaf and Ghaida Muttashar
Abdulsahib,
2019.
An
adaptive
intelligent alarm system for wireless
sensor network. Indonesian Journal of
Electrical Engineering and Computer
Science, Vol. 15, No. 1, July 2019, pp.
142~147
Shilpa, Arun Bhatia, (2016).'' Review on
Edge
Detection
Techniques
for
Pharmaceutical Drugs'', International
Journal for Scientific Research &
Development, Vol. 4, Issue 01.
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
Ranjith Pillai R, Suraj Gupta, Hirdesh
Raj. (2017). Automatic tablet blister
sorting system using Image Processing,
Journal of Chemical and Pharmaceutical
Sciences, Volume 10 Issue 1, March.
Thomas, L., & Mili, L. (2007). A robust
GM-estimator
for
the
automated
detection of external defects on barked
hardwood logs and stems. IEEE
Transactions on Signal Processing,
55(7), 3568-3576.
Lundström, J., & Verikas, A. (2010,
July). Detecting halftone dots for offset
print quality assessment using soft
computing. In Fuzzy Systems (FUZZ),
2010 IEEE International Conference on
(pp. 1-7). IEEE.
Fujiwara, N., Onda, T., & Niwakawa, M.
(1998, October). Three-dimensional
circle detection and location of pipe
joints for bin-picking tasks. In Intelligent
Robots
and
Systems,
1998.
Proceedings.,
1998
IEEE/RSJ
International Conference on (Vol. 2, pp.
1216-1221). IEEE.
Peckinpaugh, S. H., & Holyer, R. J.
(1994). Circle detection for extracting
eddy size and position from satellite
imagery of the ocean. IEEE Transactions
on Geoscience and remote sensing,
32(2), 267-273.
Chiang, A. S., Liu, P. T., Lien, Y. P.,
Huang, Y. L., & Hsieh, T. Y. (2010,
October).
Automatic
detection
of
antinuclear autoantibodies cells in
indirect immunofluorescence images. In
Biomedical Engineering and Informatics
(BMEI),
2010
3rd
International
Conference on (Vol. 1, pp. 137-140).
IEEE.
Zhang, H., Liang, C., & Wang, Y. (2011,
July). Chord midpoint randomized hough
transform
for
the
cell
image
segmentation. In Cross Strait QuadRegional Radio Science and Wireless
Technology
Conference
(CSQRWC),
2011 (Vol. 2, pp. 1446-1450). IEEE.
Kimme, C., Ballard, D., & Sklansky, J.
(1975). Finding circles by an array of
accumulators. Communications of the
ACM, 18(2), 120-122.
Fontoura Costa, L. D., & Cesar Jr, M.
(2001). R.: Shape Analysis and
Classification. Theory and Practice, CRC
Press, Boca Raton, FL.
[17] Osamah
Ibrahim
Khalaf,
Ghaida
Muttashar Abdulsahib and Muayed
Sadik, 2018. A Modified Algorithm for
Improving Lifetime WSN. Journal of
Engineering and Applied Sciences, 13:
9277-9282
[18] Young, D., Glasbey, C. A., Gray, A. J., &
Martin, N. J. (1995). Identification and
sizing of cells in microscope images by
template matching and edge detection.
[19] Lewis, J. P. (1995, May). Fast template
matching. In Vision interface (Vol. 95,
No. 120123, pp. 15-19).
[20] Jia, L. Q., Liu, H. M., Wang, Z. H., &
Chen, H. (2011, July). An effective nonHT circle detection for centers and radii.
In Machine Learning and Cybernetics
(ICMLC), 2011 International Conference
on (Vol. 2, pp. 814-818). IEEE.
[21] Canny, J. (1986). A computational
approach to edge detection. IEEE
Transactions on pattern analysis and
machine intelligence, (6), 679-698.
[22] Yuen, H. K., Princen, J., Illingworth, J.,
& Kittler, J. (1990). Comparative study
of Hough transform methods for circle
finding. Image and vision computing,
8(1), 71-77.
[23] Shang, F., Liu, J., Zhang, X., & Tian, D.
(2009, March). An improved circle
detection method based on right
triangles inscribed in a circle. In
Computer Science and Information
Engineering, 2009 WRI World Congress
on (Vol. 6, pp. 382-387). IEEE.
[24] Di, S. F. W. F. T., & Zhihui, Z. (2008). A
Method for Circle Detection Based on
Right Triangles Inscribed in a Circle .Acta
Optica Sinica, 4, 028.
[25] Hough, P. V. (1962). Method and means
for recognizing complex patterns (No.
US 3069654).
[26] Osamah Ibrahim Khalaf, Bayan Mahdi
Sabbar''An overview on wireless sensor
networks and finding optimal location of
node'',Periodicals of Engineering and
Natural Sciences, Vol 7, No 3 (2019)
[27] Ogudo, K.A.; Muwawa Jean Nestor, D.;
Ibrahim Khalaf, O.; Daei Kasmaei, H. A
Device Performance and Data Analytics
Concept for Smartphones’ IoT Services
and Machine-Type Communication in
Cellular Networks. Symmetry 2019, 11,
593.
ARTÍCULO
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
530
REVISTA AUS 26-2 / Saif S Al-jaboriy et al.,/ DOI:10.4206/aus.2019.n26.2.60/ www.ausrevista.com/ editor@ausrevista.com
ARTÍCULO
[28] Le, T., & Duan, Y. (2016, September).
Circle detection on images by line
segment and circle completeness. In
Image Processing (ICIP), 2016 IEEE
International Conference on (pp. 36483652). IEEE.
[29] Osamah Ibrahem Khalaf, Ghaidaa
Muttasher et al., “Improving video
Transmission
Over
Heterogeneous
Network by Using ARQ and FEC Error
Correction Algorithm“, vol. 30, no.8,
pp.24-27, Nov 2015
531
[30] Smith, M. L., & Stamp, R. J. (2000).
Automated
inspection
of
textured
ceramic tiles. Computers in Industry,
43(1), 73-82.
[31] GHAIDA MUTTASHAR ABDULSAHIB and
OSAMAH IBRAHIM KHALAF, 2018. AN
IMPROVED
ALGORITHM
TO
FIRE
DETECTION IN FOREST BY USING
WIRELESS
SENSOR
NETWORKS.International Journal of Civil
Engineering & Technology (IJCIET) Scopus
Indexed.Volume:9,Issue:11,Pages:369377.
[32] Segnini, S., Dejmek, P., & Öste, R.
(1999). A low cost video technique for
colour measurement of potato chips.
LWT-Food Science and Technology,
32(4), 216-222.
[33] Abou-Taleb, H. A., & Sallam, A. T. M.
(2008). On-line fabric defect detection
and full control in a circular knitting
machine. AUTEX Research Journal, 8(1).
[34] Convery, S., Lunney, T., Hashim, A., &
McGinnity, M. (1994). Automated fabric
inspection. International Journal of
Clothing Science and Technology, 6(5),
15-19.
[35] Perez, R., Silvestre, J., & Munoz, J.
(2004). Defect detection in repetitive
fabric
patterns.
Proceeding
of
Visualization,
Imaging and
Image
Processing, September, 6-8.
[36] Tajeripour, F., Kabir, E., & Sheikhi, A.
(2007). Fabric defect detection using
modified local binary patterns. EURASIP
Journal
on
Advances
in
Signal
Processing, 2008(1), 783898.
[37] Ucar, N., & ErtuฤŸrul, S. (2007).
Prediction of fuzz fibers on fabric surface
by using neural network and regression
analysis. Fibres & Textiles in Eastern
Europe, (2 (61)), 58-61.
View publication stats
[38] Yang, X., Pang, G., & Yung, N. (2005).
Robust fabric defect detection and
classification using multiple adaptive
wavelets. IEE Proceedings-Vision, Image
and Signal Processing, 152(6), 715-723.
[39] Zhi, Y. X., Pang, G. K., & Yung, N. H. C.
(2001). Fabric defect detection using
adaptive wavelet. In Acoustics, Speech,
and
Signal
Processing,2001.Proceedings.(ICASSP'0
1). 2001 IEEE International Conference
on (Vol. 6, pp. 3697-3700). IEEE.
[40] Hatcher, D. W., Symons, S. J., &
Manivannan, U. (2004). Developments
in the use of image analysis for the
assessment
of
oriental
noodle
appearance and colour. Journal of Food
Engineering, 61(1), 109-117.
[41] Brzakovic, D., Beck, H., & Sufi, N.
(1990). An approach to defect detection
in materials characterized by complex
textures. Pattern Recognition, 23(1-2),
99-107.
[42] Kauppinen, H., Rautio, H., & Silvén, O.
(1999, June). Non-segmenting defect
detection and SOM based classification
for surface inspection using color vision.
In SPIE Conference on Polarization and
Color Techniques in Industrial Inspection
(Vol. 3826, pp. 270-280).
Download