Microc alcification Segmentation Using Modified U-net Segmentation Network from Mammogram Images

Breast cancer is the most common aggressive cancer in women while the early detection of this cancer can reduce the aggressiveness. But it is challenging to identify breast cancer features such as microcalcification from mammogram images by the human eye because of its size and appearance. Therefore, the automatic detection of micro-calcification is essential for diagnosis and proper treatment. This work introduces an automated approach and segments any micro-calcification in the mammogram images. At first, the preprocessing applications of images are applied to enhance the image. After that, the breast region is segmented from the pectoral region. The suspicious regions are detected using fuzzy C-means clustering algorithm and divided them into negative and positive patches. This procedure eliminates the manual labelling of the region of interest. The positive patches which contain microcalcification pixels are taken to train a modified U-net segmentation network. Finally, the trained network is utilised to segment the micro-calcification area automatically from the mammogram images. This process can help as an assistant to the radiologist for early diagnosis and increase the segmentation accuracy of the micro-calcification regions. The proposed system is trained up with a Digital Database for Screening Mammography (DDSM), which is prepared by the University of South Florida, USA. We obtain 98.5% F-measure and 97.8% Dice score respectively. Besides, Jaccard index is 97.4%. The average accuracy of the proposed method is 98.2% which provides better performance than state-of-the-art methods. This work can be embedded with the real-time mammography system. (c) 2019 The Author. Production and hosting by Elsevier B.V. on behalf of King Saud University. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).