Praniti Golhar, Siddhivinayak Kulkarni
MIT College of Engineering, Pune
recent years, Ulcerative Colitis (UC) is most common and severe type of cancer
and to detect its stages and severity is most challenging job. Ulcerative
Colitis (UC) is a common intestinal complication which causes polyps in the
rectum which may develop further into cancer.
UC causes deaths of about half a million people every year. Colonoscopy
images are obtained by process called colonoscopy. There are various shapes of
polyps at various stages. Detection of stages of UC is impossible by naked
eyes. In an optical colonoscopy course, the endoscopist looks for colon polyps.
Hyperplastic polyp is benign lesion; adenomatous polyp is likely to become
cancerous. Hence, it is common practice to remove all of the identified polyps
and send them to subsequent histological analysis. So we are proposing a
technique in which detection of infectious area can be done with help of
medical image processing. So location of infectious area, blood clotting can be
detected with help of color impact. Shape and texture detection is used for
checking type of polyp. The technique can be useful in designing the treatment
of UC and will help in prevention of disease.
2.1 MEDICAL IMAGE PROCESSING
medical image processing, it is a process where images from interior of body
get scanned, selected and then get analyzed to check internal complications of
human body. To do this analysis, a database is formed which will be under
analysis and then use to identify abnormalities.
original data will be remain as it is during analysis. Medical images will be
great advantage for physicians to search the abnormalities in image very quickly.
2.2 ULCERATIVE COLITIS
Colitis is a Inflammatory bowel disease, which causes inflammation in ulcer,
digestive track. Ulcerative Colitis leads to many deaths in world as well as in
India which is more than a million per year. Treatment can comfort, but this
condition can’t be healed. for detecting stage of disease, there is requirement
for a medical diagnosis Lab tests or imaging always required Chronic. Ulcerative
colitis is detected in innermost lining of the large intestine (colon) and rectum.
There are various stages from mild to severe. Because of ulcerative colitis,
chances of having colon cancer will increase a lot. Symptoms for this disease
are rectal bleeding, bloody diarrhea, abdominal cramps and pain. Treatment
includes medication and surgery.
1. Colonoscopy Images.
disease has most important 4 stages and the last of them leads to cancer. The
affected abnormal part has different textures in various stages. Since these
patterns are difficult to detect with naked eyes by the doctor, PIT pattern and
Texture Recognition techniques are used. The most important algorithms used are
related to various feature extraction techniques.
2. Ages affected in UC.
is technique where CCD camera or a fiber optic camera on a flexible tube passed
through the anus. With help of this doctor can suspected colorectal cancer
lesions, polyps smaller than 1 mm. Once polyps are removed, microscope can
determine if they are cancerous or not. It is very long process to make polyp
cancerous which may be of span of 15 years.
The colonoscopy is
performed by a doctor experienced in the procedure and lasts approximately
30-60 minutes. Medications will be given into your vein to make you
feel relaxed and drowsy. You will be asked to lie on your left side on the
examining table. During a colonoscopy, the doctor uses a colonoscope, a
long, flexible, tubular instrument about 1/2-inch in diameter that transmits an
image of the lining of the colon so the doctor can examine it for any
STAGES OF ULCERATIVE COLITIS
Pit pattern: Pit pattern is
useful in checking stage of disease with help of colonoscopy, from where
severity can be checked. So it is ideal to detect stage during colonoscopy to
make process very fast.
Figure 4. Pit Pattern
2.5 IMPORTANCE OF THE RESEARCH
Development of polyp to cancer is very long, 10 to 15 year.
Chances of cancer to spread all over body, then chances for
life will e decrease.
Risk factors – overweight or obese, physically inactive,
diet, Smoking, heavy alcohol use
Colourectal cancer risk factors you cannot change : Being
older, personal and family history, Having type 2 diabetes
Factors with unclear effects on colourectal cancer risk :
Night shift work, Previous treatment for certain cancers.
cancer is largest death causing disease in world. So to analyzing this disease
medical treatment which is suggested is colonoscopy 15. Kudo et al. 5 had
discussed about patterns which are being produced by colonoscopy results. They
divided pit patterns into seven principal types: (1) normal round pit; (2)
small round pit; (3) small asteroid pit; (4) large asteroid pit; (5) oval pit;
(6) gyrus-like pit; and (7) non-pit. So further distinguish done by authors 9
16 14 on this Kudo’s pit patterns.
Pablo Mesejo et al.9 discuss about
classification of colonoscopy videos are classified in 3 classes as Adenoma,
Hyperplastic and serrated adenoma. Here machine learning algorithms are used,
it will help clinicians in virtual biopsy of hyperplastic, serrated adenoma and
adenoma. So technique is developed for diagnosis gastrointestinal lesions from
regular colonoscopy. So by using this technique systematic biopsy for suspected
hyperplastic tissues also 3D shape features improves classification accuracy.
Also many techniques analyses about
NBI images 16 3 7 17 for classification. Toru Tamaki et al. 16
divides NBI images into 3 classes as Type A, B, C3. For classification local
recognition method as Bag-of-Words is introduced along with SVM classifier.
Local features are considered and checks result for recognition checking. Hao
Chun Wang et al.3 classifies according to Classification of Regional Feature
(CoRF) which is extension of sparse matrix and vector quantization for feature
detection and segmentation. Mineo Iwatate et al. 7 discuss about efficiency and
magnifying colonoscopy with NBI image to detect, histological predict and estimation
of depth of early rectal cancer. Here, NBI International Colourectal Endoscopic
(NICE) classification is introduced here. Yasushi Sano et al. 17 discusses
about the work by Japan NBI Expert Team (JNET) where they discussed about Sano,
Hiroshima, Showa, And Jike Classifications Based on The Findings Of NBI
Magnifying Endoscopy. Also they discussed about Universal NBI Magnifying
Endoscopic Classification of Colorectal Tumors: Japan NBI Expert Team (JNET)
Classification which is universal solution which has overcome the problems
raised by previous methods. The JNET classification combining previous
classifications to give common diagnostic criteria to promote academic progress
For detection of polyp from
colonoscopy 4 14 various techniques are introduced. Ju Lynn Ong et al. 4
gives idea about features of image like geometric feature, colon wall
extraction constrction of probability density functions(PDF), comparison of
shape distribution. Here K-L divergence used for comparison between PDF for
specific image and previous database. Yuan Shen and Christopher L. Wyatt 14
uses feature extraction method for Computer Aided polyp Detection (CAPD) on
basis of principal curvature, Gaussian curvature and mean curvature, shape
index, curvedness, maximum and minimum polyp radius etc. Principle Component
Analysis is used with wrapper method.
For feature extraction of images
various techniques12613 used for extracting features. Adegoke,
B.O. et al. 1 surveyed about medical
image feature extraction. They researched about CBIR (Content- Based Image
Retrieval). The algorithms used in these systems are commonly divided into
three types as Extraction, Selection and Classification. Different available medical image feature
extraction had been studied in this paper. G.Nagarajan et al.2 proposes minimum description length principle based
genetic algorithm (GA) approach for the selection of dimensionality reduced set
of features. There are 3 phases are developed as for the extraction of the
features are Texton based contour gradient extraction algorithm, Intrinsic
pattern extraction algorithm and modified shift invariant feature
transformation algorithm. For second phase to identify the potential feature
vector GA based feature selection is done, with help of “Branch and Bound
Algorithm” and “Artificial Bee Colony Algorithm”. To improve the presentation
of the hybrid content based medical image retrieval system, feedback method is
implemented. For this algorithm they used techniques such as Intrinsic pattern
extraction algorithm using PCA. The branch and bound algorithm is used to give
reduced feature vector. M.VASANTHA et al. 6 researched about breast cancer
where proposes an image classifier to classify the mammogram images. For
preprocessing they used low pass filter to remove noise. In this Work intensity
histogram features and Gray Level Co-Occurrence Matrix(GLCM) features are
Extracted. For classification we used J48 classifier, a decision tree
classifier based on C4.5, from WEKA to train and test the features. Seyyid
Ahmed Medjahed et al.13 has done a
comparative Study of Feature Extraction Methods in Images
Classification, in which they had discussed about Feature Extraction Techniques
and classifiers on the Cal-tech 101 image dataset. classification accuracy
rate, Precision, Recall, F_measure, G_mean, AUC and the Roc Curve are used to
check the performance.
Similarly some techniques have been
developed for x-ray results1012 which also helpful for feature extraction.
Randa Hassan Ziedan et al.10 proposes
a technique for classification of x-ray images, where they discussed about
feature extraction techniques GLCM, LBP, Canny and BoW. Also a comparative
study for this techniques for x-ray images. Seyyed Mohammad Mohammadi et
al.12 also proposed shaped texture feature extraction for x-ray images. In
research Novel Shape Texture feature is proposed with help of histogram adjustment,
Noise removal, Edge and boundary extraction, phase congruency computation,
Gabor Transformation, shape-texture feature extraction with help of classifiers
as Euclidean Distance, PNN( probabilistic Neural network) and SVM.
Rathore et al. 11 discusses about colon cancer detection techniques as region
based segmentation methodologies, classification and segmentation, graph based
techniques, automated diagnosis system etc. Similarly Shiva K Ratuapli et
al.15 gives idea about post colonoscopy follow-up with help of previous
screening and surveillance colonoscopy. Mohammad Sohrabinia et al. 8 uses
different image analysis and processing methods in order to extract information
content needed to update large scale maps.
3.1 Problem Statement
To detect state/ stage of lesion, also image related more
information detection as by colour analysis (redness) or pattern formation such
type of data reporting generation are major goals.
To make a system to detect actual pattern or shape of lesion
from image also stage of Ulcerative Colitis disease.
4.2 Proposed Method
Figure 5. Proposed System Architecture
In this project, we are proposing a
method for polyp detection.
Red color filtering –
Red colour filtering will be used, which will help to detect
maximum swelling or blood clotting area. This swelling and blood clotting is
considered as initial phase of Ulcerative Colitis.
It can be considered by clinicians as most infected area.
RoI selection –
From loaded image, Region of Interest is selected.
Shape/Texture detection –
Edge detection and shape detection is done.
For edge detection, we have analyzed canny edge detector and
sobel edge detection which will give edge of RoI.
Classifier will be useful in classifying Shape/Texture
according to classes.
Final result –
Final result will be combination of red colour filtering and
4.3.1 Canny Edge Detector
algorithm runs in 5 separate steps:
Smoothing: Blurring of the image to remove noise.
Finding gradients: The edges should be marked where the gradients of the image
has large magnitudes.
Non-maximum suppression: Here edges which are detected having gradient value
are suppressed to sharp edges, but local maxima are exception.
Double thresholding: Potential edges are determined by thresholding.
Edge tracking by hysteresis: Final edges are determined by suppressing all
edges that are not connected to a very certain (strong) edge.
4.3.2 Sobel Edge Detector
6. Sobel Edge Detector
4.3.3 Red colour filtering
7. Red Colour Filtering
Figure 9 is taken as sample test case image on
which all mentioned algorithms which will be applied.
Red colour filtering
Algorithm 4.3.3 is implemented on test image (Figure 8),
which is considered as Stage 1. And result of applying Red Colour Filtering
algorithm is shown as below.
Figure 9. Red Colour Filtering
Algorithm 4.3.1 and 4.3.2 are implemented on test image
(Figure 8), which is considered as part of Stage 2. And result of applying Canny
edge Detector and Sobel Edge Detector algorithm is shown as below in Figure 10
and 11 respectively.
Canny Edge Detector
10. Canny edge detection
Sobel Edge Detctor
Figure 11. Sobel Edge Detection
help of this project, we are trying to take advantage and make a project for a
social cause which will surely help doctors with less expertise and in remote
areas to detect the stage and the probable infections in the colon image. It
will help doctors to detect the abnormalities which are not able to be seen by
the naked eye.
proposed techniques will surely help doctor to get faster result so that time
which is spent on post colonoscopy will be avoided. Also, in rural area or
unprivileged area where lack of technology is challenge then this technology
can give idea about disease severeness.
Praniti Golhar would like to thank Prof.Dr.
Siddhivinayak Kulkarni for his guidance, Prof. Deepali Jawale for her help in
this work. We would like to thank all teaching and non-teaching staff for their
support. We would like to take this opportunity to thank Head of Department
Dr.Bharati Dixit and Principal Dr. Anil Hiwale. We would also like to thank the
institute for providing the required facilities, Internet access and important
B.O, Ola, B.O. and Omotayo, M.E, “Review
of Feature Selection Methods in Medical Image Processing”, IOSR Journal of
Engineering (IOSRJEN), Vol. 04, Issue 01 (January. 2014), ||V4|| PP 01-05
G.Nagarajan, R.I.Minu, B Muthukumar, V.Vedanarayanan &
S.D.Sundarsingh, “Hybrid Genetic Algorithm for Medical Image Feature Extraction
and selection”, In Procedia Computer Science 85 ( 2016 ), 455 – 462
Hao Chun Wang, Wei Ming Chen, Yen Pin Lin, Wei Chih Shen,
“Tumor Detecting in Colonoscopic Narrow-Band Imaging Data”, In 2012 IEEE
International Symposium on Intelligent Signal Processing and Communication
Systems (ISPACS 2012) November 4-7, 2012, 564 – 568.
Ju Lynn Ong, Abd-Krim Seghouane, Kevin Osborn,” Polyp
Detection In Ct Colonography Based On Shape CharacteristicsAnd Kullback-Leibler
Divergence”, 978-1-4244-2003-2/08/$25.00 ©2008 IEEE
Kudo, S., Hirota, S., Nakajima, T., Hosobe, S., Kusaka, H.,
Kobayashi, T., Himori, M., Yagyuu, A.,
1994, “Colourectal tumours and pit pattern”. In Journal of Clinical Pathology
M.VASANTHA, DR.V.SUBBIAH BHARATHI, R.DHAMODHARAN, “Medical
Image Feature, Extraction, Selection And Classification”, International Journal
of Engineering Science and Technology, Vol. 2(6), 2010, 2071-2076
Mineo Iwatate, Taro Ikumoto, Santa Hattori, Wataru Sano,
Yasushi Sano, and Takahiro Fujimori, “NBI and NBI Combined with Magnifying
Colonoscopy”, In Hindawi Publishing Corporation, Diagnostic and Therapeutic
Endoscopy, Volume 2012, Article ID 173269, 11 pages, doi:10.1155/2012/173269
Mohammad Sohrabinia, Saeid Sadeghian, Dadfar Manavi,
“Application of Image Processing and Image Analysis Methods ForLarge Scale Map
Revision”, In the International Archives of the Photogrammetry, Remote Sensing
and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008,
Pablo Mesejo, Daniel Pizarro, Armand Abergel, Olivier
Rouquette, Sylvain Beorchia, et al.”Computer-Aided Classification of
Gastrointestinal Lesions in Regular Colonoscopy”. In IEEE Transactions on
Medical Imaging, Institute of Electrical and Electronics Engineers, 2016,
35(9), pp.2051 – 2063.
Hassan Ziedan, Mohamed A. Mead, Ghada S. Eltawel, ” Selecting the Appropriate
Feature Extraction Techniques for Automatic Medical Images Classification”, In International Journal of
Emerging Engineering Research and Technology Volume 4, Issue 5, May 2016, PP
Saima Rathore, Mutawarra Hussain, Ahmad Ali, Asifullah Khan,
“A Recent Survey on Colon Cancer Detection Techniques”,
Seyyed Mohammad Mohammadi, Mohammad Sadegh Helfroush, Kamran kazemi,”Novel Shape
Texture Feature Extraction for medical X-Ray image classification” , In
International journal of Innovative Computing, Information and Control, Volume
8, Number 1(B), January 2012, pp 659-676
Ahmed Medjahed, “A Comparative Study of Feature Extraction Methods in Images
Classification”, In I.J. Image, Graphics and Signal Processing,
2015, 3, 16-23
Shen, Y. and Wyatt, C. L., “Open implementation of feature
extraction methods for computer aided polyp detection with principal component
analysis,” in 11th International Conference on Medical Image Computing and
Computer Assisted Intervention, 141-147 (2008).
Shiva K. Ratuapli, Suryakanth R. Gurudu, Mary A. Atia,Michael
D. Crowell, Sarah B. Umar, M. Edwyn Harrison, Jonathan A. Leighton, and
Francisco C. Ramirez, “Post colonoscopy Follow-up Recommendations: Comparison
with and without Use of Polyp Pathology”, In Hindawi
Publishing Corporation, Diagnostic and erapeutic Endoscopy, Volume 2014, Article
ID 683491, 7 pages, http://dx.doi.org/10.1155/2014/683491
Toru Tamaki, Junki Yoshimuta, Misato Kawakami, Bisser
Raytchev, Kazufumi Kaneda, Shigeto Yoshida, Yoshito Takemura, Keiichi Onji, Rie
Miyaki, Shinji Tanaka, “Computer-Aided Colorectal Tumor Classification in NBI
Endoscopy Using Local Features”, In Medical Image Analysis, December 17, 2013
Yasushi Sano, Shinji Tanaka, Shin-ei Kudo, Shoichi Saito,
Takahisa Matsuda, Yoshiki Wada, Takahiro Fujii, Hiroaki Ikematsu, Toshio
Uraoka, Nozomu Kobayashi, Hisashi Nakamura, Kinichi Hotta, Takahiro Horimatsu,
Naoto Sakamoto, Kuang-I Fu, Osamu Tsuruta, Hiroshi Kawano, Hiroshi Kashida,
Yoji Takeuchi, Hirohisa Machida, Toshihiro Kusaka, Naohisa Yoshida, Ichiro
Hirata, Takeshi Terai, Hiro-o Yamano, Kazuhiro Kaneko, Takeshi Nakajima, Taku
Sakamoto, Yuichiro Yamaguchi, Naoto Tamai, Naoko Nakano, Nana Hayashi, Shiro
Oka, Mineo Iwatate, Hideki Ishikawa,Yoshitaka Murakami, Shigeaki Yoshida and
Yutaka Saito, “Narrow-band imaging (NBI) magnifying endoscopic classification
of colorectal tumors proposed by the Japan NBI Expert Team”, Digestive
Endoscopy 2016; 28: 526–533