ISSN 2249-6343 International Journal of Computer Technology and Electronics Engineering (IJCTEE) Volume 2, Issue 2
An Efficient Diagnosis of Kidney Images Using Association Rules Jicksy Susan Jose, R.Sivakami, N. Uma Maheswari, R.Venkatesh
computer output would be for radiologists in their diagnosis, how to quantify the benefits of the computer output on their diagnosis. Many different types of CAD schemes are being developed for detection of various lesions in medical images. Association rule mining is a well-known data mining technique that aims to find interesting patterns in very large databases. CAD is a relatively young interdisciplinary technology combining elements of artificial intelligence and digital image processing with radiological image processing.
Abstract- This paper is focusing to develop a method based on association rule-mining to enhance the diagnosis of ultrasound kidney images. This is to implement a computer-aided decision support system for an automated diagnosis and classification of kidney images. The method uses association rule mining to analyze the medical images and automatically generates suggestions of diagnosis. It combines automatically extracted low-level features from images with high-level knowledge given by a specialist in order to suggest the diagnosing. Association rule mining is a well-known data mining technique that aims to find interesting patterns in very large databases. The proposed method distinguishes three kidney categories namely normal, cortical cyst and medical renal. The segmented US kidney images are taken for processing. Then feature extraction process is applied to extract the features and from that extracted features only the relevant features are selected. So feature selection and discretization process is done on the extracted features that reduce the mining complexity. The association rules are generated based on the selected features. The proposed method uses the algorithm Bayesian Classifier which is a new associative classifier based on Bayesian theorem. This classifies the given image with high values of accuracy to suggest a diagnosis.
US kidney images The images that are used for the analysis are acquired from two types of scanning systems, ATL HDI 5000 curvilinear probe with transducer frequency of 3–6 MHz and Wipro GE LOGIC 400 curvilinear probe with transducer frequency of 3–5 MHz. As the sonographic evaluation is made based on the distribution of echogenity that reflects tissue characteristics, for better echo visualization the longitudinal cross section of kidney is taken to include renal sinus, medulla and cortex regions as suggested by the experts. This also ensures better visual interpretation of the normal and diseased kidney. The transducer frequency is fixed at 4 MHz. For performing the analysis the images o both right and left kidneys are considered. The necessary clinical informations of the training images are obtained. To make the testing process unbiased these details are veiled for the test images.
Index Terms— US kidney image, CAD, Image mining, Feature Extraction, Bayesian classification.
I. INTRODUCTION
Image mining
Computer-aided diagnosis (CAD) is a procedure in medicine that assists doctors in the interpretation of medical images. A computer-aided diagnosis system can be used to assist the physician’s work and t used in medical care is becoming of high importance and a priority for much research in hospitals and medical centers. For the development of a successful CAD scheme it is necessary not only to develop computer algorithms, but also to investigate how useful the
Image mining deals with the extraction of knowledge, image data relationship, or other patterns not explicitly stored in the images. It uses methods from computer vision, image processing, image retrieval, data mining, machine learning, database, and artificial intelligence. Rule mining has been applied to large image databases. There are two main approaches. The first approach is to mine from large collections of images alone and the second approach is to mine from the combined collections of images and associated alphanumeric data.
Manuscript received April 15, 2012.
Jicksy Susan Jose, PG Student Computer Science & Engineering PSNA College of Engineering and Technology,Dindigul,India. R.Sivakami,Associate Professor, MCA Dept, SONA College of Technology,Salem,India. Dr.N. Uma Maheswari, Professor, Computer Science & Engineering, PSNA College of Engineering and Technology,Dindigul,India,
[email protected]. Dr.R.Venkatesh, Professor, Information & Technology, PSNA College of Engineering and Technology,Dindigul,India.
Medical imaging is the technique and process used to create images of the human body (or parts and function thereof) for clinical purposes(medical procedures seeking to reveal, diagnose or examine disease) or medical science (including the study of normal anatomy and physiology). Although imaging of removed organs and tissues can be performed for medical reasons, such procedures are not usually referred to as medical imaging, but rather are a part of pathology.
14
ISSN 2249-6343 International Journal of Computer Technology and Electronics Engineering (IJCTEE) Volume 2, Issue 2 R. Agrawal et al. [5] in this paper, they propose a method based on association rule-mining to enhance the diagnosis of medical images (mammograms). It combines low-level features automatically extracted from images and high-level knowledge from specialists to search for patterns. Our method analyzes medical imagesn and automatically generates suggestions of diagnoses employing mining of association rules. The suggestions of diagnosis are used to accelerate the image analysis performed by specialists as well as to provide them an alternative to work on. The proposed method uses two new algorithms, PreSAGe and HiCARe. The PreSAGe algorithm combines, in a single step, feature selection and discretization, and reduces the mining complexity. Experiments performed on PreSAGe show that this algorithm is highly suitable to perform feature selection and Medical ultrasonography discretization in medical images. HiCARe is a new Diagnostic sonography (ultrasonography) is an associative classifier. The HiCARe algorithm has an ultrasound-based diagnostic imaging technique used for important property that makes it unique: it assigns multiple visualizing subcutaneous body structures keywords per image to suggest a diagnosis with high values of including tendons, muscles, joints, vessels and internal accuracy. Our method was applied to real datasets, and the organs for possible pathology or lesions. Obstetric results show high sensitivity (up to 95%) and accuracy (up to sonography is commonly used during pregnancy and is 92%), allowing us to claim that the use of association rules is widely recognized by the public. In physics, the term a powerful means to assist in the diagnosing task. P. G. Foschi et al.[8], in this paper it presents an "ultrasound" applies to all sound waves with a frequency automated recognition system for the MRI image using the above the audible range of human hearing, about 20 kHz. The neuro fuzzy logic. Texture features are used in the training of frequencies used in diagnostic ultrasound are typically the neuro-fuzzy model. Cooccurrence matrices at different between 2 and 18 MHz. directions are calculated and Grey Level Co-occurrence Medical ultrasonography uses high Matrix (GLCM) features are extracted from the matrices. It is frequency broadband sound waves in the megahertz range observed that the system result in better classification during that are reflected by tissue to varying degrees to produce (up the recognition process. The considerable iteration time and to 3D) images. This is commonly associated with imaging the accuracy level is found to be about 50-60% improved in the fetus in pregnant women. Uses of ultrasound are much recognition compared to the existing neuro classifier. Y. Liu et al [9], in this paper a new method for image broader, however. Other important uses include imaging the retrieval using high level semantic features is proposed. It is abdominal organs, heart, breast, muscles, tendons, arteries based on extraction of low level color, shape and texture and veins. While it may provide less anatomical detail than characteristics and their conversion into high level semantic techniques such as CT or MRI, it has several advantages features using fuzzy production rules, derived with the help of which make it ideal in numerous situations, in particular that it an image mining technique. Dempster-Shafer theory of studies the function of moving structures in real-time and evidence is applied to obtain a list of structures containing contain spectacles that can be used in elastography. information for the image high level semantic features. II. RELATED WORK Johannes Itten theory is adopted for acquiring high level color R. Agrawal et al.[1],they developed a novel CAD system features. The main advantage of this method is the possibility based on fuzzy support vector machine to automatically of retrieval using high level image semantic features. After the detect and classify mass using ultrasound (US) images. Breast full system realization we will be able to obtain statistic cancer can be treated most effectively when detected in its characteristics about the usefulness of the suggested method. early stage. Due to the superiority to mammography in its In this paper we propose a method which can ability to detect focal abnormalities in the dense breasts of overcome most shortcomings of the aforementioned methods adolescent women, sonography has become an important which are discussed. This method is based on association rule adjunct to mammography in breast cancer detection and has mining to enhance the diagnosis of kidney images. been especially useful in distinguishing cysts from solid The rest of this paper is organized as follows. Section tumours. The experimental results show that the proposed III describes the proposed method and the algorithms used for system greatly improves the five objective measurements and its implementation. Section IV describes about the measures the area (Az) under the ROC curve compared with those of used for the performance evaluation. Finally the paper is other classification methods, and radiologist assessments, and concluded with section V which gives a brief summary about the proposed approach will be very valuable for breast cancer the paper and the future enhancement. control.
15
ISSN 2249-6343 International Journal of Computer Technology and Electronics Engineering (IJCTEE) Volume 2, Issue 2 III.
PROPOSED METHOD
Histogram Equalization Histogram equalization is the technique by which the dynamic range of the histogram of an image is increased. Histogram equalization assigns the intensity values of pixels in the input image such that the output image contains a uniform distribution of intensities. It improves contrast and the goal of histogram equalization is to obtain a uniform histogram. This technique can be used on a whole image or just on a part of an image. Histogram equalization redistributes intensity distributions. If the histogram of any image has many peaks and valleys, it will still have peaks and valley after equalization, but peaks and valley will be shifted. Because of this, "spreading" is a better term than "flattening" to describe histogram equalization. In histogram equalization, each pixel is assigned a new intensity value based on its previous intensity level. B. Feature Extraction Feature extraction is the process of transforming the input data into the set of features to be analyzed. The feature extracted should be carefully chosen to be representative for all the relevant information from the input images in order to achieve good result from the next stages of the analysis process. Features are based on co-occurrence matrices of the texture information. All textural features are derived from the spatial gray-level dependence (SGLD) matrices, which are two-dimensional histograms. An element of the SGLDθ matrix P(i, j,d, θ) is defined as the joint probability of the gray levels i and j separated by distance d and along direction θ . In order to simplify the computational complexity, the values of θ are often given as 0◦, 45◦ , 90◦ , and 135◦, and the distance d is often defined as the Manhattan or city block distance. This procedure is detailed in [15]. The process of feature extraction in our work is mainly divided into two steps: The first step is to identify the regions of interest, and the second to extract analytical data from those regions. The analysis of images to find the regions that is affected by certain diseases is a very important process, since the effect of a certain diseases is concentrated at a specific area, so the identification of this area is critical to get accurate result from the whole classification process. For this purpose we use a MATLAB [11] toolbox that is widely used for this process. The tool box implements multiple algorithms for component analysis that is contributed by medical image specialists. We will extract the following kinds of features.
This method combines the low level features automatically extracted from images with high level knowledge given by the specialist in order to suggest the diagnosis of a new kidney image. This is to implement a computer-aided decision support system for an automated diagnosis and classification of kidney images. The proposed system is divided into mainly, i) the training phase and ii) the test phase. Algorithm: Input - training images and a test image Output - kidney category classification 1. Input image is preprocessed. 2. Extract the required features. 3. Relevant features are extracted through feature selection process. 4. Execute PreSAGe algorithm. 5. Generate association rules. 6. Classify the image based on generated association rules. A. Image Pre-Processing Since most of the real life data is noisy, inconsistent and incomplete, preprocessing becomes necessary. The cropping operation can be performed to remove the background, and image enhancement can be done to increase the dynamic range of chosen features so that they can be detected easily. The histogram equalization can be used to enhance the contrast within the soft tissue of the brain images and also hybrid median filtering technique can be used to improve the image quality. To remove the noise present in the image median filters are used. Good texture feature extraction can be done by increasing the dynamic range of gray-levels of the image. Sometimes these images may be low contrast, so in order to enhance the entire image, pre-processing is done.
Fig 3. US kidney images. a) Normal image of male with age 38 years. b) Medical renal diseases image of male with age 45 years and c) cortical polycystic disease image of female with age 51 years.
16
ISSN 2249-6343 International Journal of Computer Technology and Electronics Engineering (IJCTEE) Volume 2, Issue 2 For the present application, the estimation of texture Fractal features Fractal concept is useful to represent a statistical features along this specific angles shows lacking ability in quality of roughness and self-similarity at different scales of identifying the kidney category, hence the analytical natural surfaces and/or curves. The fractal dimensions as expression meant for describing the features are modified to geometric features have become popular in modeling image include extended definition of angles at 0°, 30°, 60°, 90°, properties. Intuitively, the degree of roughness of the image 120° and 150°. Haralick proposed 14 texture measures that texture is proportional to the fractal dimension. The definition can be extracted from Pkl(i, j). Here five texture features of the fractal dimension is similar to the Hausdorff dimension. namely energy (E), entropy (H), correlation (C), inertia (In) Informally, self-similar objects with parameters N, the and homogeneity (L) are calculated. number of similar pieces, and s, the magnification factor, are C. Feature Selection described by N = s f , where f is known as the Hausdorff Feature selection process can be performed using dimension. In this work, five features are extracted based on PreSAGe algorithm. PreSAGe (Pre-processing Solution for the fractal dimensions: Association rule Generation) is a novel supervised algorithm • fp1 = fractal dimension of the ROI. that performs feature selection. Before explaining the • fp2 = fractal dimension of the suspicious area. PreSAGe algorithm the following definitions are necessary. • fp3 = fractal dimension of the ROI excluding Definition 1: Class is the most important keyword given by a the suspicious area. specialist for the diagnosis. • fp4 = fp2/fp1. Definition 2: Cut Points are the limits of an interval of values • fp5 = fractal dimension of the contour of the taken. suspicious area. Definition 3: Majority Class is the most frequent class of an interval. Histogram-based features Definition 4: An instance Ii belongs to an interval if its value The shape of the histogram provides many clues to is given between two consecutive cut points. Two input describe the characteristics of the image. Six statistic features thresholds used by this algorithm: extracted from the histogram are mean, variance, skewness, • minperint: it will restricts the minimal number kurtosis, energy, and entropy. The mean is the average of occurrences of the majority class allowed in intensity level whereas the variance implies the variation of an interval; intensities around the mean. The skewness indicates whether • mintofuse: it will restricts the minimum the histogram is symmetric about the mean. The histogram is occupancy of the majority class in an interval. symmetrical if the skewness is zero. Otherwise, it is skewed The following conditions are to be satisfied for determining above the mean if the skewness is positive, and skewed below the cutpoints. the mean if the skewness is negative. The kurtosis is a • Condition 1: The class label of the current measure of whether the data are peaked or flat relative to a instance, Ii, i ≥1 is different from the class label normal distribution. Data with high kurtosis tend to have a of the previous instance, i.e., ci !=ci -1. distinct peak near the mean, declining rather rapidly, and Condition 1 generates too many cut points, especially having heavy tails. Data with low kurtosis tend to have a flat when working with noisy data. The larger the number of cut top near the mean. Entropy is a measure of how much disorder points, the larger the number of intervals. It will generate a in a system. number of irrelevant rules. Hence, it is necessary to keep the number of cut points small and, consequently, generating a First feature extraction technique small number of items as well. The unnecessary cut points can be removed by using the next two steps: The first order gray level statistical features mean • Condition 2: The number of occurrences of (M1), dispersion (M2), variance (M3), average energy (M4), the majority class in a particular interval must skewness (M5), kurtosis (M6), median (M7) and mode (M8) be equal or greater than the minperint are estimated for preprocessed US kidney images. threshold. • Condition 3: The middle cut point of two consecutive Second feature extraction technique intervals and is removed if Mk=Mk+1. This algorithm is also used to select the most relevant features The spatial gray level dependence matrix (SGLDM) is from the extracted features, according to the following one of the most widely used techniques for statistical texture criterion: description. All known visually distinct texture pairs can be • Criterion 1: The features that generate the discriminated using this method. These statistical features of smallest number of cut points are selected as second order are computed in two steps. The first step the most relevant ones delivers the cooccurrence matrices containing the element A threshold valreduct is used to state the percentage of Pkl(i, j). Each (i,j)th entry of the matrices represents the reduction of the original number of features compared to the probability of going from pixel with gray level (i) to another extracted features. The whole process can be speedup by with a gray level (j) under predefined angles. Usually for reducing the irrelevant features by using the PreSAGe statistical texture analysis these angles are defined at 0°, 45°, algorithm. 90° and 135°.
17
ISSN 2249-6343 International Journal of Computer Technology and Electronics Engineering (IJCTEE) Volume 2, Issue 2 c {i1, …, ik-1, i’k-1}; // join f1 and f2 Ck Ck {c}; for each (k-1)-subset s of c do if (s Fk-1) then delete c from Ck; // prune end end return Ck; E. Bayesian Classification
PreSAGe Algorithm Input: image feature vector F, image classes C, minperint, mintofuse and valreduct thresholds. Output: processed feature vector V 1. For each feature f F do 2. Sort f values 3. For each transaction I, create an instance Ii of the form ci,fi where ci C Uses condition 1 to create the vector U of cut point up 4. 5. End for 6. For each up U do 7. Remove up according to condition 2 8. Remove up according to condition 3 Uf 9. Save the remaining cut points in a vector 10. End for 11. Rank the features f according to number of cut points in Uf 12. Select the 1-valreduct * |F| features that generate the least number of cut points. 13. Write the selected features discretized in V 14. Return V
The Bayesian classifier is a statistical classifier that performs probabilistic prediction i.e. predicts class membership probabilities. The foundation is based on Bayes’theorem. A simple Bayesian classifier, naïve Bayesian classifier has comparable performance with decision tree and selected neural network classifiers. Let X be a data sample. Let H be the hypothesis that X belongs to class C. Classification is to determine P(H/X) the probability that the hypothesis holds given the observed data sample X. P(H) (prior probability), the initial probability, P(X) probability that sample data is observed, P(X|H) (posteriori probability), the probability of observing the sample X, given that the hypothesis holds. Given training data X, posteriori probability of a hypothesis H, P(H|X), follows the Bayes theorem. P (H/X) = P(X/H) P (H) P(X) The association rules are given to the Bayesian classifier and will classify the kidney images as normal, medical renal or cortical cyst. Based on the generated association rules the images are classified.
D. Association Rule Generation The association rules are generated from the selected features. It is an important data mining model studied extensively by the database and data mining community. Initially used for Market Basket Analysis to find how items purchased by customers are related. Find all rules that satisfy the user-specified minimum support and minimum confidence. Based on the relevant features selected the association rules are generated using Apriori algorithm [5].
IV. PERFORMANCE EVALUATION
Algorithm Apriori(T) C1 init-pass(T); F1 {f | f C1, f.count/n minsup}; // n: no. of transactions in T for (k = 2; Fk-1 ; k++) do Ck candidate-gen(Fk-1); for each transaction t T do for each candidate c Ck do if c is contained in t then c.count++; end end Fk {c Ck | c.count/n minsup} end return F k Fk; Function candidate-gen(Fk-1) Ck ; forall f1, f2 Fk-1 with f1 = {i1, … , ik-2, ik-1} and f2 = {i1, … , ik-2, i’k-1} and ik-1 < i’k-1 do
Fig 4. Input kidney image
The actual efficiency obtained gives the accuracy of the systems in identifying the kidney categories. But often mostly used empirical measure accuracy does not account all correct or incorrect labels of different categories. The classification performance without focusing on a different category may be the most general way of comparing the decision support systems. The number of labels belonging to one category may be substantially lower than the overall number of labels.
18
ISSN 2249-6343 International Journal of Computer Technology and Electronics Engineering (IJCTEE) Volume 2, Issue 2 REFERENCES [1] R. Agrawal, T. Imielinski, and A. N. Swami, ―Mining association rules between sets of items in large databases,‖ in Proc. 1993 ACMSIGMOD Int. Conf. Manage. Data - SIGMOD 93 SIGMOD 93,Washington, DC, 1993, pp. 207–216. [2] K. Beyer, J. Godstein, R. Ramakrishnan, and U. Shaft, ―When is nearest neighbor meaningful?,‖ in Proc. Int. Conf. Database Theor. (ICDT), 1999, pp. 217–235. [3] N. R. Mudigonda and R. M. Rangayyan, ―Detection of breast masses in mammograms by density slicing and texture flow-field analysis,‖ IEEE Trans. Med. Imag., vol. 20, no. 12, pp. 1215–1227, Dec. 2001. [4] R. S. Kazemzadeh and K. Sartipi, ―Incorporating data mining applications into clinical guidelines,‖ in Proc. 19th IEEE Int. Symp. Computer- Based Med. Syst., Salt Lake City, UT, 2006, pp. 321–328. [5] R. Agrawal and R. Srikant, ―Fast algorithms for mining association rules,‖ in Proc. Int.Conf. VLDB, Santiago, Chile, 1994, pp. 487– 499. [6] M. Klemettinen, H. Mannila, P. Ronkainen, H. Toivonen, and A. Verkamo, ―Finding interesting rules from large sets of discovered association rules,‖ in Proc. 3rd Int. Conf. Inf. Knowledge Manage. - CIKM 94 CIKM 94, Gaithersburg, MD, 1994, pp. 401–407. [7] P. Haiwei, J. Li, and Z. Wei, ―Medical image clustering for intelligent decision support,‖ in Proc. 2005 27th Annu. Int. Conf. IEEE Eng. Medicine Biol. Soc., Shanghai, China, 2005, pp. 3308–3311. [8] P. G. Foschi, D. Kolippakkam, H. Liu, and A. Mandvikar, ―Feature extraction for image mining,‖ in Proc. 8th Int. Workshop Multimedia Inf. Syst., Tempe, AZ, 2002, pp. 103–109. [9] Y. Liu, N. A. Lazar, W. E. Rothfus, M. Buzoianu, and T. Kanade,―Classification- driven feature space reduction for semanticbased neuroimage retrieval,‖ in VISIM Workshop: Inf. Retrieval Exploration Large Collections Med. Images, A.W.M. Smeulders and S. Ghebreab, Eds., Utrecht, The Netherlands, 2001, pp. 4–4. [10]W. Hsu, M. L. Lee, and K. G. Goh, ―Image mining in IRIS: Integrated retinal information system,‖ in Proc. ACM SIGMOD Int. Conf. Manage. Data, Dallas, TX, 2000, pp. 593– 1593. [11] C. Ordonez and E. Omiecinski, ―Discovering association rules based on image content,‖ in Proc. IEEE Forum ADL, 1999, pp. 38–49. [12]O. R. Zaiane, M.-L. Antonie, and A.Coman, ―Mammography classification by an association rule-based classifier,‖ in Proc. 3rd Int. Workshop Multimedia Data Min. (MDM/KDD’2002), Edmonton, AB,Canada, 2002, pp. 62–69. [13] A. Olukunle and S. Ehikioya, ―A fast algorithm for mining association rules in medical image data,‖ in Proc. IEEE CCECE2002 Canadian Conf. Electr. Comput. Eng. Conf. Proc. (Cat No 02CH37373) CCECE-02, 2002, pp. 1181– 1187.
Fig 5. Histogram equalization of kidney image
Fig 6. Fractal dimension of kidney image Within a set of categories (NR, MRD and CC) there is a category of interest (usually positive) and other two categories that are combined to misclassification category. The performance of the proposed system can be calculated based on specificity, accuracy and sensibility. V. CONCLUSION The diagnosing of diseases as well as to enhance the health care of patients, the increasing use of image exams in the last 25 years has greatly improved. As the volume of images has grown at a fast pace, the radiologists have more and more images to manually analyze. Thus the process of diagnosing becomes tiresome and consequently more susceptible to errors. In order to avoid such bottleneck we go for computer aided diagnosing. In this paper we presented a method which is based on association rules and it can be integrated into a CAD. Our approach is divided into four major steps: pre-processing, feature extraction and selection, association rule generation, and generation of diagnosis suggestions from classifier. The results are applied to real databases and the proposed system achieves high sensitivity and accuracy for diagnosing. This brings more confidence to the diagnosing process. The feature extraction step can be improved to obtain more representative features for the future works. Since feature extraction is the first step for diagnosing the medical images. It will also be interesting to investigate the applicability of our proposed method for other medical images.
19
ISSN 2249-6343 International Journal of Computer Technology and Electronics Engineering (IJCTEE) Volume 2, Issue 2 [14] R. M. Haralick, K. Shanmugam, and I.Distein, ―Textural features for image classification,‖ IEEE Trans. Syst., Man, Cybern., vol. SMC-3, pp. 610–621, 1973. [15] J. C. Felipe, A. J. M. Traina, and C. Traina, ―Retrieval by content of medical images using texture for tissue identification,‖ in Proc. 16th IEEE Symp. Computer-Based Med. Systems. CBMS 2003, New York, 2003, pp. 175–180. [16] K. Kira and L. A. Rendell, ―A practical approach for feature selection,‖ in Proc. 9th Int. Conf. Mach. Learning, Aberdeen, Scotland, 1992, pp. 249–256. [17] R. Quinlan, C4.5: Programs For Machine Learning. San Mateo, CA: Morgan Kaufmann, 1993. [18] G. H. John and P. Langley, Estimating Continuous Distributions in Bayesian Classifiers. San Mateo, CA: Morgan Kaufmann, 1995, pp. 338–345. [19] Jiang, Y., Nishikawa, R. M., Schmidt, R. A., Metz, C. E., Giger, M. L., Doi, K., Improving breast cancer diagnosis with computeraided diagnosis. Acad. Radiol. 6:22–33, 1999. [20] Aoyama, M., Li, Q., Katsuragawa, S., MacMahon, H., Doi, K., Automated computerized scheme for distinction between benign and malignant solitary pulmonary nodules on chest images. Med. Phys. 29(5):701–708, 2002.
Dr.N.Uma Maheswari. received her M.E in Computer Science & Engineering from the Madras University, Chennai, India in 2002 and Ph.D. in Information & Communication engineering in 2011 at Anna University, Chennai. Currently, she is working as Professor in the Department of Computer Science & Engineering at the P.S.N.A. College of Engineering & Technology, Dindigul, India. Her current research interests include Image processing, Artificial Intelligence, Speech Processing. She has published 8 papers in International journals, 1 paper in national journals and 10 papers international conferences and 12 papers in national conferences. She is a recognized Ph.D. supervisor under Anna University of Technology in the area of Image processing; Cloud computing, Network security & Networks.
Dr.R.Venkatesh received his ME in Computer Science and Engineering from Anna University, Chennai, India in 2007 and Ph.D. in Computer Science & Engineering in 2010. Currently he is working as Professor in the Department of Information Technology in PSNA College of Engineering & Technology, Dindigul, in India. His current research interests include Artificial intelligence, Neural Networks, Soft computing , Network Security and Networks. He has published 10 papers in International journals, 1 paper in national journal and 12 papers in international conferences and 10 papers in national conferences.
Ms.Jicksy Susan Jose is pursuing her ME in Computer science & Engineering at PSNA College of engineering and technology, Dindigul. Her research area includes Image processing, Pattern recognition and Artificial intelligence.
Ms.R.Sivakami received her M.E in Computer Science & Engineering from Anna University, Chennai, India in 2008 and pursuing Ph.D. in Information & Communication engineering at Anna University, Coimbatore. Currently, she is working as Associate Professor in the Department of Master of Computer Applications at SONA College of Technology, Salem, India. Her current research interests include Image processing, Computer Networks, Mobile Computing ,Network Security and Artificial Intelligence. She has published 8 papers in International Conferences including IEEE Conference Proceedings, 2 papers in International journals and 12 papers in national conferences.
20