Abstract: A method for automatically segmenting a volume-of-interest representative of a subject's breast from a three-dimensional magnetic resonance image is provided. The three-dimensional image may include a plurality of spatially contiguous two-dimensional images. The image is converted to a monogenic signal, which is analyzed to determine locations in the image that correspond to maximal phase congruency in the monogenic signal. The orientation of each of these locations is determined and used along with the locations to estimate a boundary surface of the volume-of-interest. The estimated surface may be used to segment the image directly, or to generate a surface model, such as a statistical shape model, that is used to segment the image. This provided method is robust to segmenting the subject's breast, even at the chest-wall boundary in images with lower contrast-to-noise ratio between breast tissue and tissues in and around the chest wall.

Abstract: A method for automatically segmenting a volume-of-interest representative of a subject's breast from a three-dimensional magnetic resonance image is provided. The three-dimensional image may include a plurality of spatially contiguous two-dimensional images. The image is converted to a monogenic signal, which is analyzed to determine locations in the image that correspond to maximal phase congruency in the monogenic signal. The orientation of each of these locations is determined and used along with the locations to estimate a boundary surface of the volume-of-interest. The estimated surface may be used to segment the image directly, or to generate a surface model, such as a statistical shape model, that is used to segment the image. This provided method is robust to segmenting the subject's breast, even at the chest-wall boundary in images with lower contrast-to-noise ratio between breast tissue and tissues in and around the chest wall.