SAMIR KUMAR BANDYOPADHYAY [IJESAT] INTERNATIONAL JOURNAL OF ENGINEERING SCIENCE & ADVANCED TECHNOLOGY

ISSN: 22503676
Volume - 2, Issue - 1, 34 37

EDGE DETECTION FROM CT IMAGES OF LUNG

Abstract
Lung Cancer detection is important since if doctor may detect it earlier then it is easy to them to diagnose and may be possible to recover the disease in the early stage. It is one of the first growing diseases now a day in the world. This paper provides a method using Computer Aided Diagnosis System (CAD) for detection of edges from CT images of lung for detection of diseases.

Index Terms: CT Images, Edge Detection and Neighborhood filter --------------------------------------------------------------------- *** -----------------------------------------------------------------------1. INTRODUCTION
Image classification has become an important problem in several domains and has found its various applications in various spheres. One of the main areas where this field has gained importance is medical imaging. X-rays have been used for years for detecting internal anatomy and anomalies. During the past decades, image processing has been able to diagnose pathologies based on X-rays [2-3]. In this paper lung images of 50 individuals were studied for edge detection from images. The frontal view of X-ray is shown in figure 1. shape of a Gaussian (`bell-shaped') hump. The idea of Gaussian smoothing is to use this 2-D distribution as a `point-spread' function, and this is achieved by convolution. It is known that White noise is one of the most common problems in image processing [2]. For a high resolution photo texture will be represented by a large group of pixels. The main idea of any neighborhood filter is to calculate pixel weights depending on how similar their colors are. The input image is a normal RGB image. The grey scale image from RGB contains noises and can be removed by using Gaussian smoothing from the extracted lung image. In this paper we have proposed a method for detecting edge rather than using one of the known methods of edge detection.

2. LITERATURE REVIEW
Yamomoto et al., proposed image processing for computeraided diagnosis of lung cancer by CT (LSCT) [1]. LSCT is the newly built mobile-type CT scanner mainly for the purpose of mass screening of lung cancer. In this new LSCT technique, one main complexity is the raise in the image information to around 30 slices per person from 1 X-ray film. In the process detection of edge is needed to display image information to the doctor for identify abnormalities. Kanazawa et al., [4] used Computer aided diagnosis system for lung cancer based on helical CT images. This method reduces the time complexity and increases the diagnosis confidence. This method consists of an analysis stage and a diagnosis stage. It is known that appearance of small nodules is one of the early signs of lung TB and cancer. Nodule pixels are often brighter than the surrounding areas but in some cases, the

Figure 1 Frontal View of lung X- Ray [7] The accurate interpretation of edge detection is one of most important step to go further for detection of abnormalities. Image de-noising algorithms may be the oldest in image processing. Many methods, regardless of implementation, share the same basic idea noise reduction through image blurring. Blurring can be done locally, as in the Gaussian smoothing model or in anisotropic filtering by calculating the variations of an image. The Gaussian smoothing operator is a 2-D convolution operator that is used to `blur' images and remove detail and noise. It uses a kernel that represents the

IJESAT | Jan-Feb 2012

SAMIR KUMAR BANDYOPADHYAY [IJESAT] INTERNATIONAL JOURNAL OF ENGINEERING SCIENCE & ADVANCED TECHNOLOGY difference in grey levels is not significant. Furthermore, ribs and pulmonary arteries, which often have higher grey levels, also contribute to the complexity of lung tissue and make some nodules undetectable. In the analysis stage, the lung and pulmonary blood vessel regions are extracted and examine the features of these regions with the help of image processing methods. In the diagnosis stage, diagnosis rules are determined according these features, and identify the tumor regions using these diagnosis rules. Researcher Yamamoto et al., [5] discussed Computer aided diagnosis system with functions to assist comparative reading for lung cancer based on helical CT image. This paper provided a new and automatic system for an early detection of lung cancer based on a computer aided diagnosis system in which all the CT images are read. Additionally, the CAD system is provided with functions to automatically identify the suspicious regions from chest CT images, and to provide the comparative reading in retrospect. Edges are usually defined as those points in an image where the gradient magnitude is maximum in the gradient direction. Ridges are commonly defined as extrema in the direction of the largest curvature. Instead of these definitions, consider structures defined by extreme in a fixed direction, for derivatives of a certain order in that same direction. In this way we obtain structures that cannot cross, because in the direction of the derivative consecutive points cannot both the extremal. This implies that such structures can never turn, i.e., a line in the direction of the derivative cannot cross a structure more than once. The order of derivative n and the choice for minima and maxima determines the nature of the detected structures. The slice matching algorithm is used in this paper for comparing every slice image of the current and previous CT scans, and an interface to show some features of the suspicious regions. Cheran et al., [6] provided computer aided diagnosis for lung CT using artificial life models. This CAD system is based on several techniques such as 3D region growing, active contour and shape models, centre of maximal balls but it can be believed that at the centre of this technique are the biological models of ant also called as artificial life models. In the starting of the process the 3D region growing algorithm is applied on the image for detecting the ribcage. After the ribcage is identified an active contour is used for building a confined area for the incoming ants that are set up to make clean and perfect restoration of the bronchial and vascular tree. Next the branches of the newly restored trees are verified to detect whether they comprise nodules or not with the help of active shape models and also for determining if there are any nodules connected to the pleura of the lungs. The next phase is to remove the trees in order to provide a cleaner technique for localizing the nodules which is performed with the help of snakes and dot enhancement techniques.

ISSN: 22503676
Volume - 2, Issue - 1, 34 37

3. PROPOSED METHOD
In this paper initially digitized image is stored as pixel in an array. First step of the process is to determine the dimension of the image and determine the middle position of image array. We then take a maximum difference threshold which is constant threshold determined by the intensity of the image. The homogeneity process is done in two passes. In the first pass the image is scanned in horizontal direction considering each row from the first row till the last row of the image. In the next pass the same process is repeated in the vertical direction by considering a column wise scanning, starting from first column till the last column of the image. The resultant image obtained after the two passes represents the lung image that has homogenous regions that is color quantized to sixteen-color in gray scale. This image forms the input image for the edge detection process. After the image was processed by the proposed homogeneity we obtain an image that is homogenous and have pixels in gray scale of sixteen level scales. This image forms the input for the next level of processing for detection of the edge, i.e. lung boundary and the lung region. It can be observed that all the lung images of our data set has well defined baseline that separates the lung image from the body and contains black pixels. We consider these features in our proposed method of edge detection. The edge detection process is done in three passes. In the first pass the scanning of image is done horizontally in the x direction i.e. in a row major order. The process starts from the leftmost pixel of the first row and traverse all pixels on the first row to reach the last pixel of the first row. Then we repeat the scanning of pixel from the next row and continue to subsequent rows till we reach the last pixel in the last row of the two dimensional image data array. A threshold value is taken for pixel intensity comparison. The threshold value changes depending on the intensity of the pixels in the surrounding region. During the scanning of pixels the algorithm compares the intensities of subsequent pixels. If a change in intensity is observed between two pixels that exceed the absolute threshold value then the algorithm marks the last pixel as edge and adds this pixel location to the horizontal edge map image. The process now continues the scanning of image vertically in the y direction i.e. in a column major order. The algorithm starts from the topmost pixel of the first column and traverse

IJESAT | Jan-Feb 2012

SAMIR KUMAR BANDYOPADHYAY [IJESAT] INTERNATIONAL JOURNAL OF ENGINEERING SCIENCE & ADVANCED TECHNOLOGY all pixels on the first column to reach the last pixel of the first column. Then we repeat the scanning of pixel from the next column and continue to subsequent columns till we reach the last pixel in the last column of the two dimensional image data array. Finally, in the last pass we merge the Horizontal Edge Map image with Vertical Edge Map image by performing a Union operation on the two image files, to obtain the Edge map of mammogram image. The edge image obtained uniquely contains single pixel continuous edge line as a result we need not perform further edge thinning process. The edge line is continuous so the process of edge linking is also not required. It can be observed from the edge map image that each region of the lung in image is clearly visible. The edge detection process has also separated the lung boundary from the exterior non lung region. With this edge detection method, the image segmentation process becomes very easy as most of the lung region is already isolated by the process. This edge detection method is simple, fast as it does not require convolution, robust and can also be implemented to other medical imaging with little modifications. It helps to detect nodule in the lung of any size. The outputs from the process are shown in figure 2 and figure 3.

ISSN: 22503676
Volume - 2, Issue - 1, 34 37

4. CONCLUSION
Detection of edges in an image is a very important step towards understanding image features. Since edges often occur at image locations representing object boundaries, edge detection is extensively used in image segmentation when images are divided into areas corresponding to different objects. This paper proposes a method using Computer Aided Diagnosis System (CAD) for detection of edges from CT images of lung for detection of diseases.

REFERENCES
[1] Yamomoto, S., Jiang, H., Matsumoto, M., Tateno, Y., Iinuma, T. and Matsumoto, T., Image Processing for Computer-Aided Diagnosis of Lung Cancer by CT (LSCT), IEEE Workshop on Applications of Computer Vision, Pp. 236 241, 1996. [2] Y. Hirano and J. Hasegawa and J. Toriwaki and H. Ohmatsu and K. Eguchi: Computer Classification of Lung Tumors from Chest CT Images According to the Types ofTissue Using 3D Extended Voronoi Diagram and CARS 2002 Computer Assisted Radiology and Surgery (Proc. CARS 2002, Paris, France and June 2002), Springer and pp.729-733,2002. [3] Y.Hirano, J.Hasegawa, J.Toriwaki, H.Ohmatsu and K.Eguchi : Extraction of tumor regions keeping boundary shape information from chest X-ray CT images and benign/malignant discrimination ,Proc. of Computer Assisted Radiology and Surgery (CARS)2001 [4] Kanazawa, K., Kubo, M. and Niki, N., Computer Aided Diagnosis System for Lung Cancer Based on Helical CT Images, International Conference on Pattern Recognition, Vol. 3, Pp. 381- 385, 1996.

Figure 2 Original CT Image

[5] Yamamoto, T., Ukai, Y., Kubo, M., Niki, N., Satou, H. and Eguchi, K., Computer Aided Diagnosis System with Functions to Assist Comparative Reading for Lung Cancer Based on Helical CT Image, International Conference on Image Processing, Vol. 1, Pp: 180 - 183, 2000. [6] Cheran, S.C. and Gargano, G., Computer Aided Diagnosis for Lung CT Using Artificial Life Models, Seventh International Symposium on Symbolic and Numeric Algorithms for Scientific Computing, 2005. [7] http://homepages.inf.ed.ac.uk/rbf/H1PR2/Kernel.htm

Figure 3 Edge Detection Output Image

IJESAT | Jan-Feb 2012

SAMIR KUMAR BANDYOPADHYAY [IJESAT] INTERNATIONAL JOURNAL OF ENGINEERING SCIENCE & ADVANCED TECHNOLOGY

ISSN: 22503676
Volume - 2, Issue - 1, 34 37

BIOGRAPHIES
Dr. Samir Kumar Bandyopadhyay, B.E., M.Tech., Ph. D (Computer Science & Engineering), C.Engg., D.Engg., FIE, FIETE,Sr. Member IEEE, currently, Professor of Computer Science & Engineering, University of Calcutta, Kolkata, India. Visiting Faculty, Dept. of Comp. Sc., Southern Illinois University, USA, MIT, California Institute of Technology, etc. His research interests include Biomedical Engg, Mobile Computing, Pattern Recognition, Graph Theory, Software Engg etc. He has 25 Years of experience at the Post-graduate and under-graduate Teaching & Research experience in the University of Calcutta. He has already got several Academic Distinctions in Degree level/Recognition/Awards from various prestigious Institutes and Organizations. He has published 300 Research papers in International & Indian Journals and 5 leading text books for Computer Science and Engineering. Dr. Bandyopadhyay is the former Registrar of University of Calcutta and West Bengal University of Technology, Kolkata, and formar Acting-VC, West Bengal University of Technology, Kolkata, India.

IJESAT | Jan-Feb 2012