• Segmentation-driven CAD workflows for mammography dominate, emphasizing automatic mass delineation, ROI localization, and boundary refinement via contour searching, active contours, multiresolution thresholding, and level-set methods to enable robust detection [2], [9], [15], [17], [3].

• Texture-based representations and quantitative descriptors drive tissue characterization across mammography and ultrasound, leveraging co-occurrence statistics, orientation-aware textures, multiresolution features, and Gabor filters to separate benign from malignant lesions [10], [12], [11], [13], [19].

• Learning-based discriminators, including neural networks, SVMs, and logistic regression, are used to automate diagnosis and prognosis from image-derived features and ROIs [2], [10], [19], [20], [8].

• Computational pathology data resources and the broader prognosis discourse shape the field, with discussions of data sharing initiatives and the promises and challenges of tissue analysis [6], [14], [5].

• Cross-modality segmentation advances extend beyond mammography into ultrasound and 3D contexts, using shape statistics, longitudinal registration, and level-set-like approaches to robustly delineate structures in breast and prostate imaging [16], [18], [8].