Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/261164047
CITATIONS READS
0 65
4 authors, including:
SEE PROFILE
All content following this page was uploaded by Kamarul HAWARI Ghazali on 04 November 2015.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
2012 International Symposium on Computer Applications and Industrial Electronics (ISCAIE 2012), December 3-4, 2012, Kota
Kinabalu Malaysia
patients. Besides the advancement of medical electronics stage, and testing stage. For this purpose, the proposed
equipment in the diagnosis of disease, there is still room approach was tested by using test data to detect sixteen
for improvement, especially when the diagnosis occurs in kinds of parasites. At the end of the test, 97.70% overall
volume or its being manually inspected by technologists. success rates were obtained [11].
The Goal of this paper is to present a viable method for III. Methodology
detecting human parasites using image processing which
evaluates five features in the body structure of the Most researchers who have extensively studied parasite
parasite. This approach makes the detection highly classification extensively focused on classification
accurate and less time consuming. Detection is made in method to detect the type of parasite. Among the famous
two to four seconds per image, which is far lower than classification tools are ANN, SVM, MCSVM and
compared to predictions done using the conventional ANFIS. Few of them are doing analysis at the features
manual process. selection or the image processing stage. In this research,
parasite images were collected as data from Hospital
II. Literature review Universiti Sains Malaysia (HUSM), Kelantan, Malaysia.
For both types of parasites, Ascaris Lumbricoides ova
In previous studies, microscopic imaging analysis has (ALO) and Trichuris trichiura ova (TTO), 100 images
been used for recognition and detection of parasites [3]. were collected. Some of them were not clear and were
The aim of the study is to classify several parasite types rejected; the remaining images were used as a database to
occurring in human beings by using morphologic features system development. The second feature of this
extracted in interactive image tools [4]. These methodology is image processing analysis, which
morphologic features were subsequently included in the consists of pre-processing, image processing, feature
computation of classification indices. The parasite types extraction, and classification. There are different methods
were then recognized by using the calculated indices [5]. for the pre-processing stage, which have been tested to
Yang et al. (2001) has proposed an automated method get the best results. Fig. 1, Among those three methods,
to detect common helminthes eggs in microscopic fecal only one method is the suitable, resulting in less error and
specimen images. For that purpose, the digital image keeping the image data from being lost. Based on the
processing techniques and artificial neural network experiment, the result showed that pre-method III was the
(ANN) classifier are used. ANN classification was best method.
performed by two sub classifiers. Most ANNs classify Pre-method III removed the noise totally, since the
given patterns according to their geometric clustering in median filter was used twice. Contrast enhancement was
the pattern space [1]. An ANN identified the parasite performed, so the brightest and the darkest pixel values
species based on their characteristics. In his study, four are estimated. Sharpness method was used to sharpen the
features were selected based on three morphological outlines of all objects. Then, by using edge detection to
characteristics representing shape, shell smoothness, and clarify the edge of all objects inside the image, it goes to
size. the processing stage. The differences between the three
Sengur and Turkoglu (2004) have proposed a method of pre-processing methods are in the use of some processing
recognizing parasite eggs using pattern recognition tools and sequences of those tools. For example, in pre-
technique. In their study, the feature extraction method I, it used converter images from gray images to
mechanism based on the invariant moments and artificial binary images, based on threshold images. This was not
neural network classifier was used; using that method, it useful, since it helped to lose the main features of the
was reported that more than 90% true recognition rates image, although this method is widely used in image
were obtained [6][7]. Finally, classification was carried processing, which is why, (the pre-method I) wasnt
out: a specific pattern was assigned to a specific class successful. It wasnt successful in this study because it
according to its characteristic features, selected for it deals with parasite eggs and very small details, which
[8,9]. must be recognized in the detection stage. Additionally,
Esin Dogantekin (2008) proposed a robust technique pre- method II ignored the convertor image to prevent the
based on invariant moments adaptive network based loss of detail regarding input data, but it used edge
fuzzy inference system (ANFIS) for recognition of detection with canny filter. That is a way to detect the
human parasite eggs in microscopic images. The edge of hard sharp objects. Since the two kinds of
technique presented in the study enabled classification of parasite eggs in this study have very clear borders, using
sixteen different parasite eggs from their microscopic edge detection with canny filter was one of the main
images. This proposed recognition method included three steps in the pre-processing stage. However, using pre-
stages [10]. method II was not a solution for all images that have been
Derya Avci and Varol Asaf (2009) proposed an expert tested because some eggs have hard background artifacts,
diagnosis system for classifying human parasite eggs which means their borders wouldnt be very clear and
based on multi-class SVM, which was based on invariant sharp. While in pre-method III, most problems in pre-
moments and multi-class support vector machine method I & II have been solved for better results by using
(MCSVM) for classification of human parasite eggs in edge enhancement (sharpness) and edge detection (canny
microscopic images. This method consists of pre-
processing stage, feature extraction stage, classification
197
2012 International Symposium on Computer Applications and Industrial Electronics (ISCAIE 2012), December 3-4, 2012, Kota
Kinabalu Malaysia
filter). From the beginning, it twice used a median filter obtained. Table I shows feature values records of some
to reduce artifacts and remove as much noise as it can. parasites (for ALO and TTO).
TABLE I
FEATURE VALUE FOR SOME ALO AND TTO
Figure 2.The algorithm of the TLCM Method used in this studying Based on the literature review, the most common
approach is to use pattern recognition at the classification
In the feature extraction stage, five features have been stage. However, in this study, a different method of
selected based on three characteristics. Those detection is proposed. Using that method, focus on the
characteristics represent shape, shell smoothness, and image processing stage and characteristics get the best
size (these parameters are section area, objects length, results, so simple classification tools can be used to train
objects width, length of boundary, and roundness). technicians to more accurately classify parasites. The
These five features are obtained by image processing classification was based on the values which are recorded
tools on MATLAB environment. The five parameters are as shown in Table I. The result comparison table shown
the major features that are able to detect the parasite eggs in Table II is between three different pre-methods; that
and each parameter have limited maximum and minimum information summarizes the methodology for three sets
values, which are called range. These ranges are recorded of results that clearly demonstrate that pre-method III is
by experimenting and testing until the best results are the best method for the detection of helminthes eggs,
which a success rate of 93% for ALO with 7% error, and
198
2012 International Symposium on Computer Applications and Industrial Electronics (ISCAIE 2012), December 3-4, 2012, Kota
Kinabalu Malaysia
94% for TTO with 6% error. Compared to pre-method I images as an input data which are going to be captured
& II, which are less successful in terms of results, Pre- from patients specimen in a microscope and each single
method III is the technique that reaches the best results in image is going to be analyzed separately. So if the system
detection of parasites. As shown in Table 4.1, pre-method finds one parasite (which its very seldom that can detect
I and II was 73% and 81% successful at ALO detection; one parasite only in one patients specimen) or finds 10s
and 75%, 84% of TTO successful detection. The main parasites then the system will decide that patient is
point of the comparison in Table II is to show that using infected with ALO or TTO or with both worms.
unsharp masking [12] and Canny edge detection [13]
together is more powerful to detect ALO and TTO types V. CONCLUSION
of parasitic worms. Notably, Table II doesnt show any
comparison with other studies in same field. The systematic diagnosis of two types of human
Table III clearly shows that 93% of successful detection parasitic worms, Ascaris Lumbricoides ova (ALO) and
is achieved for ALO and 94% of successful detecting was Trichuris trichiura ova (TTO) used image processing
reached for TTO. This result can be reliably reached technique with MATLAB toolbox. The algorithm
using the recommended methods in this study, which is presented in this paper is limited to the basic diagnosis of
ready to be used to detect ALO & TTO automatically. these kinds of parasitic worms. This technique is capable
of detecting the existence of ALO and TTO parasites
TABLE II within few seconds per image; thus, this method can
RESULTS COMPARISON
replace the conventional methods of detecting worms in
biomedical applications and this program can count the
Parasite Name Ascaris Trichuris
number of parasites which have been detected for each
lumbricoides ova trichiura ova
(ALO) (TTO) single patient. Further a user friendly environment is also
provided with simple GUI which helps to produce results
No. of Images 100 100 consuming lesser in man power and in time as well
No. of Images reached between 1-2 seconds per image as compare to
with successful 93 94
detect conventional methods.
No. of Images
with 7 6 (a)
unsuccessful
detect
% Ratio of Error 7% 6%
% Ratio of
Successful 93 % 94 %
detect for overall
images
TABLE III
RESULTS RECORDES (b)
199
2012 International Symposium on Computer Applications and Industrial Electronics (ISCAIE 2012), December 3-4, 2012, Kota
Kinabalu Malaysia
(a) [7] Avci & Akpolat, Speech recognition using a wavelet packet
adaptive network based fuzzy inference system. Expert Systems
with Applications, 2006, 31(3), 495503.; Avci, Turkoglu, &
Poyraz, (2005) Intelligent target recognition based on wavelet
adaptive network based fuzzy inference system. Lecture Notes in
Computer Science (Vol. 3522/2005). Springer-Verlag..
[11] Derya Avci & Asaf Varol. An expert diagnosis system for
classification of human parasite eggs based on multi-class SVM,
2009.
REFENRENCES
[1] Yang, Park, Kim, Choi, & Chai. Automatic identification of
human helminth eggs on microscopic fecal specimens using
digital image processing and an artificial neural network. IEEE
Transcations on Biomedicine, 48(6), 2001, 718730.
[2] Ash & Orihel,. Atlas of Human Parasitology (3rd ed.). Chicago,
IL: ASCP, 1990.
[6] Sengur, A., & Turkoglu, I. Parasite egg cell classification using
invariant moments. In: Proceedings of 4th international
symposium on intelligent manufacturing systems, 2004, (pp. 98
106).
200
2012 International Symposium on Computer Applications and Industrial Electronics (ISCAIE 2012), December 3-4, 2012, Kota
Kinabalu Malaysia