Abstract

A novel algorithm was developed to facilitate segmentation of the articular cartilage of the knee from magnetic resonance (MR) image data. The new approach uses a model of the bone and cartilage elements of the knee as a template. The operator initializes the template by aligning certain bony anatomic landmarks. The inner cartilage surface (ICS) is determined through a standard seed growing segmentation approach; the bone elements of the template are then fit to the ICS through an elastic registration technique. The deformation fields thus obtained are subsequently applied to the appropriate cartilage templates. The deformed templates and ICS normals, in conjunction with a gaussian probability distribution function and the subject image data, are used to produce the final estimate of the cartilage surfaces.

This content is only available via PDF.
You do not currently have access to this content.