Mass Segmentation Using One Single View

Segmentation using a single mammographic image relies on the fact that pixels inside a mass have different characteristics from the other pixels within the breast area. The characteristics used can be simply related to intensity values and to (local) texture measures, or otherwise related to morphological features like the size and margins of the mass, as Figure  schematizes. In addition, some approaches take the distribution of spicules associated with masses into account. Both aspects can be treated independently or sequentially. The columns of Table  classify proposed methods according to the characteristics they use for the segmentation, while the properties used in the optional subsequent classification processes (benign/malignant discrimination) are not taken into account.

Segmentation techniques can be divided into supervised and unsupervised approaches. Supervised segmentation, also known as model-based segmentation, relies on prior knowledge about the to be segmented and optional background regions. The prior information is used to determine if specific regions are present within an image or not. Unsupervised segmentation consists of partitioning the image into a set of regions which are distinct and uniform with respect to specific properties, such as grey-level, texture or colour. Classical approaches to solve unsupervised segmentation are divided in three major groups [56]:

• Region-based methods, which divide the image into homogeneous and spatially connected regions.
• Contour-based methods, which rely on finding the boundaries of regions.
• Clustering methods, which group together those pixels having similar properties and might result in non-connected regions.

The rows of Table  show the reviewed works arranged according to their classifications. In subsequent subsections, a detailed description of the methods in each category (region-based, contour-based, clustering, and model-based methods) is given.