International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169 Volume: 3 Issue: 2 606 - 612 _______________________________________________________________________________________________ 606 IJRITCC | February 2015, Available @ http://www.ijritcc.org _______________________________________________________________________________________ Depth Segmentation Method for Cancer Detection in Mammography Images Parvathy S Kumar ECE Dept MCET Trivandrum, Kerala sasikumarparvathy@gmail.com Priyanka Prabhakar ECE Dept Regional Centre IHRD Trivandrum, Kerala priya27982@gmail,.com Aneesh R P ECE Dept Regional Centre IHRD Trivandrum, Kerala aneeshprakkulam@gmail.com Abstract— Breast cancer detection remains a subject matter of intense and also a stream that will create a path for numerous debates. Mammography has long been the mainstay of breast cancer detection and is the only screening test proven to reduce mortality. Computer-aided diagnosis (CAD) systems have the potential to assist radiologists in the early detection of cancer. Many techniques were introduced based on SVM classifier, spatial and frequency domain, active contour method, k-NN clustering method but these methods have so many disadvantages on the SNR ratio, efficiency etc. The quality of detection of cancer cells is dependent with the segmentation of the mammography image. Here a new method is proposed for segmentation. This algorithm focuses to segment the image depth wise and also coloured based segmentation is implemented. Here the feature identification and detection of malignant and benign cells are done more easily and also to increase the efficiency to detect the early stages of breast cancer through mammography images. In which the relative signal enhancement technique is also done for high dynamic range images. Markovian random function can be used in the depth segmentation. Markov Random Field (MRF) is used in mammography images. It is because this method can model intensity in homogeneities occurring in these images. This will be helpful to find the featured tumor Keywords- Computer aided diagnosis, KNN, Mammography, Markovian radon function SVM. __________________________________________________*****_________________________________________________ I. INTRODUCTION Cancer is a continual multiplying of cells abnormally. The cells divide uncontrollably and will grow into adjacent tissue or unfold to distant parts of the body. Carcinoma is a number one reason for cancer deaths among women in many parts of the world. Breast cancer continues to be the foremost common diagnosed cancer among women in US. In the United States, every year, approximately, 182,000 new cases of breast cancer are diagnosed and more than 46,000 women die of it Since the causes of breast cancer still remain unknown, early detection is the key to control the breast cancer . Nowadays, mammography is considered to be the most reliable imaging modality for an early detection of breast carcinomas. If the tumor is detected at an early stage, the chances of successful treatment as well as patient survival rate will increases considerably. A mammogram is essentially distinct with four levels of the intensities such as background, breast parenchyma, fat tissue and calcifications with increasing intensity. Masses develop from the epithelial and connective tissues of breasts and their densities on mammograms blend with parenchyma patterns. Presently digital mammography is the most efficient and widely used technology for early carcinoma detection. The key diagnosing elements like masses, lesions in the digital mammograms are noisy and of very low contrast. An efficient segmentation approach to detect the early disease detection of breast cancer by enhancing the images of tumour. Most of the limitations of conventional mammography can be overcome by using digital image processing. Thus, in order to improve the correct diagnosis rate of cancer the image enhancement techniques are widely used to enhance the mammogram and assist radiologists in detecting it. Some of the efficient enhancement algorithm of digital mammograms based on wavelet analysis and modified mathematical morphology. Adopt wavelet-based level dependent thresholding algorithm and modified mathematical morphology algorithm to increase the contrast in mammograms to ease extraction of suspicious regions known as regions of interest (ROIs) are used. Several segmentation techniques are used like the gradient vector flow snake (GVF Snake) with gradient map adjustment to obtain the accurate breast boundary from the rough breast boundary and an improved multi-scale morphological gradient watershed segmentation method for automatic detection of clustered microcalcification in digitized mammograms. II. LITERATURE REVIEW The design and evaluation of the imaging system Clear- PEM for positron emission mammography, under development by the PEM Consortium within the framework of the Crystal Clear Collaboration at CERN, is presented in [1]. The camera consists of two compact and planar detector heads with dimensions 16.5×14.5 cm2 for breast and axilla imaging. Low- noise integrated electronics provide signal amplification and analog multiplexing based on a new data-driven architecture. The coincidence trigger and data acquisition architecture makes extensive use of pipeline processing structures and multi-event memories for high efficiency up to a data acquisition rate of one million events/s. Experimental validation of the detection techniques, namely the basic properties of the radiation sensors and the ability to measure the depth-of-interaction of the incoming photons, are presented. In the multi-stage system [2] propose, segmentations of the breast, the nipple and the chestwall are performed, providing landmarks for the detection algorithm. Subsequently, voxel features characterizing coronal spiculation patterns, blobness, contrast, and depth are extracted. Using an ensemble of neural- network classifiers, a likelihood map indicating potential abnormality is computed. Local maxima in the likelihood map are determined and form a set of candidates in each image. These candidates are further processed in a second detection stage, which includes region segmentation,