CAD Based System for Automatic Detection & Classification of Suspicious Lesions in Mammograms

Veena, Jayakrishna V

Abstract It is very difficult to detect the breast cancer in its earlier stage as it shows symptoms gradually. So it is very important to keep track of any abnormality present in the breast. Breast cancer is the leading cause of death due to cancer among the women. Computer Aided Diagnosis (CAD) is an important tool in assisting doctors in the early detection of cancers. Earlier diagnosis of breast cancer is of great importance as far as modern medical treatments are concerned. Mammography is currently used for the reliable and early detection of cancers in breast. This paper presents, a novel algorithm for CAD based breast cancer analysis. The proposed system develops a systematic scheme to classify the breast cancer into normal, benign and malignant type. Here, AI based Training and Classification method using Back Propagation Neural Network classifier is proposed. These classification techniques find even the minute area of suspicion which can be either cancerous or fatty and this area is used for classification. Adaptive thresholding and multiresolution analysis are used for lesion detection. Multiresolution analysis provides a framework for interpreting the image information. The proposed system uses wavelets transform, coarse segmentation, fine segmentation and area Analysis. Wavelet transform is used for multi-resolution analysis. Wavelet based method is used for coarse segmentation and window based method is used for fine segmentation. For obtaining the rough localization of suspicious lesions, coarse segmentation is used. A window based adaptive thresholding is done on the resultant image to get the fine segmentation. Fine segmentation is used for accurately spotting and detecting even the fine lesions. Area analysis of white pixel is performed in fine segmented image. The proposed method is tested with images in the Mini Mammography database of the Mammographic Image Analysis Society (MIAS). Keywords: CAD, neural network classifier, multi-resolution analysis, wavelet transform, coarse segmentation, fine segmentation, mass classification

Title:	CAD Based System for Automatic Detection & Classification of Suspicious Lesions in Mammograms
Author Name:	Veena, Jayakrishna V
Abstract:	Abstract It is very difficult to detect the breast cancer in its earlier stage as it shows symptoms gradually. So it is very important to keep track of any abnormality present in the breast. Breast cancer is the leading cause of death due to cancer among the women. Computer Aided Diagnosis (CAD) is an important tool in assisting doctors in the early detection of cancers. Earlier diagnosis of breast cancer is of great importance as far as modern medical treatments are concerned. Mammography is currently used for the reliable and early detection of cancers in breast. This paper presents, a novel algorithm for CAD based breast cancer analysis. The proposed system develops a systematic scheme to classify the breast cancer into normal, benign and malignant type. Here, AI based Training and Classification method using Back Propagation Neural Network classifier is proposed. These classification techniques find even the minute area of suspicion which can be either cancerous or fatty and this area is used for classification. Adaptive thresholding and multiresolution analysis are used for lesion detection. Multiresolution analysis provides a framework for interpreting the image information. The proposed system uses wavelets transform, coarse segmentation, fine segmentation and area Analysis. Wavelet transform is used for multi-resolution analysis. Wavelet based method is used for coarse segmentation and window based method is used for fine segmentation. For obtaining the rough localization of suspicious lesions, coarse segmentation is used. A window based adaptive thresholding is done on the resultant image to get the fine segmentation. Fine segmentation is used for accurately spotting and detecting even the fine lesions. Area analysis of white pixel is performed in fine segmented image. The proposed method is tested with images in the Mini Mammography database of the Mammographic Image Analysis Society (MIAS). Keywords: CAD, neural network classifier, multi-resolution analysis, wavelet transform, coarse segmentation, fine segmentation, mass classification
Cite this article:	Veena, Jayakrishna V , " CAD Based System for Automatic Detection & Classification of Suspicious Lesions in Mammograms " , International Journal of Emerging Trends & Technology in Computer Science (IJETTCS), Volume 3, Issue 4, July - August 2014 , pp. 338-345 , ISSN 2278-6856.
Full Text [PDF] Home