A comparative study on soft computing based classification and detection algorithms of anomalous patterns from mammograms

Abstract

Breast cancer becomes apparent as one of the primary sources of deadly diseases worldwide especially among women community. The normal body cells can be transmuted into cancerous ones. However, there is confirmation that early recognition and treatment can raise the survival rate of breast cancer patients. The guidelines for early detection of breast cancer include breast self-exams, clinical breast examination, and screening mammogram. Screening mammography is the most common imaging procedure for diagnosing breast cancer, usually among women who have no complaints or symptoms of breast cancer. The goal of a screening mammogram is to detect cancer when it is still too small to be felt by a woman or her physician. newlineFor the early detection of breast cancer, many classification algorithms and automatic decision-making systems were implemented by the researchers. To interpret and recognize the pattern of the mammogram abnormality, the radiologists are using Computer Aided Diagnosis (CAD). To differentiate normal, benign, and malignant stages of cancer and extract the suspicious area, the contrast enhancement, feature selection, histogram, and Gray Level Co-occurrence Matrix (GLCM) are deployed. The best features are selected by using Correlation-based Feature Selection (CFS) methods. The statistical values such as mean, standard deviation, smoothness, angular second moment, entropy, and correlation are considered as best features in which they guaranteed the improvement of classification with fewer features dimensions. Even though many researchers are doing investigation in this area, diagnosing cancer in the early stage is an extraordinary challenge in the medical field. In the first part of the work, an automated system is implemented to classify the normal, benign and malignant breast tissues. Noises presented in the images are removed using Gaussian Mixture Model, and to improve the appearance of the image, the Contrast-Limited Adaptive Histogram Equalization (CLAHE) algorithm is used. K-Means clustering segmentation is accomplished to identify the abnormality in the mammogram. Hybrid feature extraction method is deployed to extract the features from the region of the mammogram which includes Gray Level Co-occurrence Matrix (GLCM), texture, and gradient. The features such as contrast, correlation, energy, homogeneity, global mean, uniformity, entropy, and skewness are considered as the best features that assured the improvement of classification with a reduced amount of feature dimension newline