Radiologist analyzes thousands of mammograms daily and this causes fatigue and eye strain. Out of thousands of cases, not more than 10 have breast cancer tissues and thus an abnormality might not be detected. To assist radiologists in order to detect mammographic abnormalities, computer-aided detection (CAD) algorithms are developed. This work proposes a CAD system which makes use of principal and independent component analysis (PCA and ICA). Classifiers used in this work are support vector machines (SVMs), classifier and artificial neural network (ANN) classifier. SVM or ANN identifies the suspicious regions and classifies them into the labels assigned. PCA which is a necessary pre-processing step for ICA reduces the dimensionality of the input images. ICA extracts the maximum independent features from the reduced data set and then SVM or ANN classifies the test image into normal or malignant. A comparative analysis of the classification accuracy by using ANN and SVM as classifier is done. For both these classifiers, different numbers of principal components are experimented. © 2014 WIT Press.