An Efficient Method for Automated Breast Mass Segmentation and Classification in Digital Mammograms

 Automatic detection and classification of breast masses in mammogram images are still a challenging task. Computer-aided diagnosis (CAD) systems are being developed to assist radiologists in interpreting mammograms.

 The aim of this study is to provide a novel method for automatic segmentation and classification of the masses in mammograms to assist radiologists for accurate diagnosis.

 For the purpose of efficient mass diagnosis in mammogram images, we propose an automatic scheme to perform both detection and classification. Firstly, a combination of several image enhancement algorithms that include the contrast limited adaptive histogram equalization (CLAHE), Guided image, and median filtering is investigated to enhance the breast area visual details and make the segmentation result more accurate. Secondly, the Density of Wavelet Coefficients (DWC) based on Quincunx Lifting Scheme (QLS) is proposed to find suspicious mass regions (region of interest, ROIs). Finally, the mass lesions that appear in the mammogram image are classified into four categories based on the morphological shape properties as benign, probably benign, malignant and probably malignant. The proposed method is evaluated on 1593 images from CBIS-DDSM dataset.

 The experimental results demonstrate that the proposed suspicious region localization achieves 100% sensitivity with an average of 6/4 ± 4/5 False Positive (FP) detections per image. Moreover, the results showed an overall accuracy of 85.9% and AUC of .901 for the mass classifying algorithm.

 The results showed that the proposed method suggests comparable performance to the state-of-the-art methods.