BREAST CANCER EARLY DETECTION USING X RAY IMAGES

Volume 119 No. 15 2018, 399-405 ISSN: 1314-3395 (on-line version) url: http://www.acadpubl.eu/hub/ http://www.acadpubl.eu/hub/ BREAST CANCER EARLY DETECTION USING X RAY IMAGES Kalaichelvi.K 1,Aarthi.R 2,Abinaya.M 2,Afreetha.A 2,Bavatharani.T 2 kalaish123@gmail.com aarthirajap@gmail.com abiatchaya1104@gmail.com afreethaanwar@gmail.com tbavatharanibe@gmail.com 1 Assistant Professor, Department of Electronics and Communication Engineering, 2 UG Students, Department of Electronics and Communication Engineering, V.S.B. Engineering College, Karur-639 111. Abstract-Breast cancer is the most common form of cancer among women worldwide. Early detection of breast cancer can increase treatment options and patient s survivability. In the field of radiology, X ray images are complement to explain. The expert radiologists prefer X - ray images for any specific abnormality. This paper proposes a Computer Aided Detection (CAD) system for reducing the human risks also to help the radiologist for determining whether the breast tissues as benignant or malignant or non malignant by automatic diagnosis. Using K-mean segmentation, we can partition the images. After image segmentation feature extraction and selection are the next step to reduce volume of data processed. Application of GLCM to extract texture features for motion estimation of images. Convolution Neural Network is one of the most popular techniques used to improve the classification accuracy of detection. Keywords-CNN, Computer Aided Diagnosis System, GLCM, X-rays. I.INTRODUCTION Cancer is a group of diseases that cause cells in the body to change and grow out of control. Most of the cancer cells eventually form a lump or masses called a tumor, and are named after the part of the body where the tumor originates. Breast cancer begins in breast tissues, which are made up of glands for milk production, called lobules, and the ducts that connect lobules to the nipple. Breast cancer is leading cause of cancer deaths among women. Detection and diagnosis of breast cancer in its early stage increases the chances for successful treatment and complete recovery of the patient. X ray images are currently the best available radiological technique for early detection of breast cancer. In X rays digital medical image sensing, CAD can be classified in two ways: (a) the CAD system point out the region of suspicious (ROS) and alert the expert radiologist to the need for further analysis and (b) the CAD system which can take the decision from ROS whether it is normal, benign or malignant tissue. II.LITERATURE SURVEY In this section, we summarize the most relevant existing research are: (a)ensemble Learning - It helps to improve machine learning results by combining several model. This approach allows the production of better predictive performance compared to a single model. (b)mammography Mammogram uses a machine designed to look only at breast tissue. The machine takes X rays at lower doses than usual X rays. This gives a better picture and allows less radiation to be used. 399

(c)intra-thoracic fluid volume Correct diagnosis of pleural effusion as either benign or malignant is crucial, although conventional cytological evaluation is of limited diagnostic accuracy, with relatively low sensitivity rates. (d)automated Computerized Scheme Develop a computer-aided detection method and investigate its feasibility for detection of breast cancer in automated 3D ultra sound images of women with dense breasts. Database III.MATERIALS AND METHOD: The proposed CAD system is tested and validated on the portion of Mammographic Image Analysis Society (MIAS) database. The original MIAS database is down sampled to 200 micro pixel edge, clipped and padded so that every image has size 1024*1024. The mammographic image analysis society has produced a digital mammography database. Original mammograms were carefully selected from a major centre, ensuring high quality exposure and patient positioning. Methodology Computer Aided detection or diagnosis (CAD) systems have been developed to help radiologists in order to increase diagnosis accuracy. Generally, a CAD system consists of four steps: (a). preprocessing, (b). segmentation of regions of interest, (c). feature extraction and selection, and finally (d). Classification Pre-processing Pre- processing is the most important step in the X ray image analysis due to poor captured X ray image quality. Pre-processing is very important to correct and adjust X ray images for further study and processing. There are different type of techniques are available for preprocessing. The techniques like adaptive median filter, mean filter, adaptive mean filter, histogram equalization and contrast limited adaptive histogram equalization. This techniques used to improve image quality, remove the noise, preserves the edges within an image, enhance and smoothen the image. (a)histogram Equalization The process of adjusting intensity values can be done automatically using histogram equalization. It involves transforming the intensity values so that the histogram of the output image approximately matches a specified histogram. The original image has low contrast, with most pixel values in the middle of the intensity range. histeq produces an output image with pixel values evenly distributed throughout the range. Block diagram: (b)adaptive Histogram Equalization Input images Output images preprocessing Classifier (CNN) Segmentation (K- mean Algorithm) Feature extraction (GLCM) As an alternative to using histeq, that can perform contrast limited adaptive histogram equalization (CLAHE) using the adapthisteq function. While, histeq works on the entire image, adpathisteq operates on small region in the image. After performing the equalization, adapthisteq combines neighboring tiles using bilinear interpolation to eliminate artificially induced boundaries. 400

To avoid amplifying any noise that might be present in the images, that can use adapthisteq. Segmentation Image segmentation is the partitioning of an image into separate groups. Many researchers have been done in the area of image segmentation using clustering. One of the most popular methods is K- mean clustering algorithm. It is an unsupervised area from the background. But before applying this algorithm, first enhance the image to improve the quality of the image. Group the data into K clusters where K is already defined constant value. Select K points at random as cluster centers. Assign objects to their closest cluster center according to the Euclidean distance function. Calculate the centroid or mean of all objects in each cluster. Repeat above steps until the same points are assigned to each cluster in consecutive rounds. Feature Extraction The feature extraction is to extract features from X-ray images to improve performances of the diagnosis. All features can be coarsely classified into low-level and high-level features. K- MEAN ALGORITHM START NO OF CLUSTERS K CENTROID DISTANCE OBJECTS TO CENTROIDS NO OBJECT MOVE GROUP? END Grey Level Co-occurrence Matrix (GLCM) method is a way of extracting second order statistical texture features. A GLCM is matrix where the number of rows and columns is equal to number of gray levels in the image. It has been used in several works extract the texture information of breast modules. This type of method is most widely used texture analysis method in biological imaging, due to its ability to capture the spatial dependence of gray level values within an image. Auto-correlation: GROUPING BASED ON MINIMUM DISTANCE Autc = i j i,j p i,j μ x μ y σ x σ y (1) Entropy: K-mean is relatively an efficient method. However, we need to specify the number of clusters, in advance and the final results are sensitive to initialization and often terminates at a local optimum. A practical approach is to compare the outcome of multiple runs with different K and choose the best one based on a predefined criterion. In general, a large K probably decreases the error but increases the risk of over fitting. Ent = - i j p(i, j) log(p(i, j)) (2) Sum variance: Svar = 2N g i=2 (i saver) 2 p x+y (i) (3) Sum entropy: 2N g Sent = p i=2 x+y (i) log(p x +y (i)) (4) 401

Difference variance: N g 1 i 0 Diffv = - (i μ x y ) 2 p x y (i) (5) Information measure of correlation 2: classifier as a training source. It will then produce a model to recognize the X ray image. Input: Imc2 = (1- exp[-2.0(hxy2)-ent]) 1/2 (6) Contrast: N 1 Cont = i,j =0 p i,j (i-j) 2 (7) N 1 Dissimilarity: Diss = i,j =0 p i,j i-j (8) Samples of input images from MIAS database Output: Energy: Ener = N 1 i,j =0 p(i, j) 2 (9) Cluster prominence: Clpr = i,j(i + j μ x μ y ) 4 p[i,j] (10) Cluster shade: Clsh = i,j (i + j μ x μ y ) 3 p[i,j] (11) Classifier Neural Network (NN): It plays an important role in this respect, especially in the application of breast cancer detection. One of the most popular techniques is Convolution neural network. In this paper, breast cancer detection using CNN for X-ray imaging system is proposed to classify X-ray image into normal, benign and malignant. It is aimed to speed up the diagnosis process by assisting to diagnosis and classification of breast cancer. A series of X-ray images are used to carry out preprocessing to convert a human visual image into computer visual image and adjust suitable parameter for the suitable CNN classifier. After that, all challenged images are assigned into CNN Samples of output images IV.PERFORMANCE ANALYSIS The performance of the proposed system is estimated by choosing the number of False Positive (FP), False Negative (FN), True Positive (TP) and True Negative (TN). If the image is clustered with more than one abnormality and, at least, one of them is detected by the proposed algorithm, we have taken the result as TP. All the findings outside the sense of above definition are taken as TN. TP+TN Accuracy = TP+FN+TP+FP TP Sensitivity = TP+FN TN Specificity = FP+TN 402

Comparison table of the proposed CAD system with other CAD system Methods Year Database Classifier Sensitivity (%) Specificity (%) Accuracy (%) GLCM 2018 MIAS CNN 99.1 98 98.5 PCPCET 2016 MIAS ADEWNN 98.19 97.19 97.96 PCET 2016 MIAS ADEWNN 95.10 94.35 94.64 MIAS 92.70 90.54 90.16 Rouhi et al. 2015 MLP DDSM 96.87 95.94 96.47 Dheeba etal. 2014 Clinical Database PSOWNN 94.16 92.10 93.67 V.CONCLUSION AND FUTURE WORK This paper reviewed the literature on the use of CAD systems for breast cancer detection and diagnosis in X-ray images. The main stages of CAD system include pre-processing, segmentation of ROI, feature extraction and selection, and finally classification. The input Region of Interest (ROI) is extracted automatically. The designed system attains a fair accuracy of 98.5% with 99.1% sensitivity and 98% specificity. The evaluation metrics were also reviewed for assessment of CAD system on X-ray images.. In this study we only computed mass shape features for classification. In our future work, we need to further investigate the segmentation and features in this group to improve the classification and for the proposed CAD system we plan to add texture and speculation features. References [1] Zhang, Y., Tomuro, N., Furst, J., & Raicu, D. S. (2012). Building an ensemble system for diagnosing masses in mammogram. Internatioal journal of computer assisted radiology and surgery, 7(2), 323-329 [2] Jalalian, A., Mashohor, S. B., Mahmud, H. R., Saripan, M. I. B., Ramli, A. R. B., & Karasfi, B. (2013). Computer-aided detection/diagnosis of breast cancer in mammography and ultra sound: a review. Clinical imaging, 37(3), 420-426 [3] Urooj, S., Khan, M., Ansari, A. Q., Lay- Ekuakille, A., & Salhan, A. K.(2012). Prediction of quantitative intrathoracic fluid volume to dignose pulmonary oedema using LabVIEW. Computer methods in biomechanics and biomedical engineering, 15(8), 859-864. [4] Suckling, et al., the Mammographic Image Analysis Society Digital Mammogram Data base, ExerptaMedica, International Congress Series,1069, 1994, pp. 375-378. [5]Suckling, et al., the Mammographic Image Analysis Society Digital Mammogram Data base, ExerptaMedica, International Congress Series,1069, 1994, pp. 375-378. [6]Rouhi, R., Jafari, M., Kasaei, S., & Keshavarzian, P.(2015). Benign and malignant breast tumors classification based on region growing and CNN segmentation. Expert Systems with Applications, 42(3), 990-1002. [7]Singh, S. P., & Urooj, S. (2015). Combined Rotation-and Scale-Invariant Texture Analysis Using Radon-Based Polar Complex Exponential Transform. Arabian Journal for Science and Engineering, 114. 403

404

405

406