Computer Aid Diagnostic in Mammogram Image Using SUSAN Algorithm and Hierarchical Watershed Transform
Chaimae Anibou, Mohammed Nabil Saidi, Driss Aboutajdine · Lecture notes in electrical engineering · 2016
This work is directed toward a conception of a computer aid diagnosis (CAD) system to detect suspicious area in digital mammogram and classify them into normal and abnormal. Original image is preprocessed to separate the breast region from it’s background with pectoral muscle suppression to reduce false positive rate. The suspicious regions are extracted using a modified SUSAN algorithm, followed by a function that extract dense regions, then an hierarchical watershed transforms applied to detect edges of suspicious regions. For detected edges Fourier Descriptors are computed and stored as feature vector. A support vector machine is used to classify suspicious regions into normal or abnormal. The proposed system is tested on Mini-Mias database.