Benign Calcification Detection in Mammogram Images

This paper describes an algorithm for detecting calcifications which are benign in nature. Computer Aided Detection (CAD) for breast cancer is useful for screening and for second look, because it assists the radiologist to evaluate a large number of patient cases and also to improve accuracy of cancer detection. Calcification is one of the important abnormalities which indicate cancer in breast. But they can be seen along with many benign changes too. Hence it is essential to differentiate between benign or malignant calcifications and accurately list the finding in the final mammography report for avoiding unrequired biopsy of patients. The proposed method uses a three level wavelet enhancement technique. Reconstruction of the image by scaling the detailed coefficients emphasizes the calcifications in the image. Background tissue suppression of the reconstructed image is performed with the help of morphological filtering. The microcalcifications and skin thickening regions which get delineated along with calcifications are filtered out by area thresholding and edge detectors respectively. Experimental results show that multilevel wavelet reconstruction method can be effectively used for benign calcification detection by suitable selection of wavelet type and detailed coefficient scaling factor. The algorithm gives better results with an image level true positive rate of 96.15%. This approach results in reduced number of false positives with a false positive rate of 3.85%. For evaluation of the proposed method a database of digital images in DICOM format is employed.