Curve evolution for medical image segmentation
Liana M. Lorigo, Olivier D. Faugeras, W. Eric L. Grimson · 2000
The model of geodesic curves in three dimensions is a powerful tool for image segmen-tation and also has potential for general classification tasks. We extend recent proofs on curve evolution and level set methods to a complete algorithm for the segmentation of tubular structures in volumetric images, and we apply this algorithm primarily to the segmentation of blood vessels in magnetic resonance angiography (MRA) images. This application has clear clinical benefits as automatic and semi-automatic segmen-tation techniques can save radiologists large amounts of time required for manual segmentation and can facilitate further data analysis. It was chosen both for these benefits and because the vessels provide a wonderful example of complicated 3D curves. These reasons reflect the two primary contribu-tions of this research: it addresses a challenging application that has large potential benefit to the medical community, while also providing a natural extension of previous geometric active contour models research. In this dissertation, we discuss this extension and the MRA segmentation system,