Method and system for segmentation of tubular structures using pearl strings

4 The method of claim 1, further comprising:

• receiving user inputs identifying a first region in said 2D image inside of the tubular structure and a second region in said 2D image outside of the tubular structure;
• and
• estimating a classification function that classifies a pixel in said 2D image as inside or outside of the object being segmented, based on the first and second regions, respectively.

The method of claim 4, wherein said step of estimating a classification function comprises:

• estimating first and second probability densities for the first and second regions, respectively, using non-parametric density estimation based on the pixel intensities in the first and second regions, wherein a pixel in said 2D image is determined to be inside or outside of the tubular structure based on the first and second probability densities.

The method of claim 6, wherein steps (a) and (b) are performed subject to constraints on a distance between the center of a current pearl and the center of a previous pearl, where the constraints are defined at least in terms of the respective radii of the current and previous pearls.

The method of claim 6, wherein said step of generating an ordered series of pearls further comprises, for each pearl:

• (d) determining whether the pearl intersects with any previous pearls, excluding a number of pearls immediately preceding the pearl in the ordered series of pearls;
• and
• (e) if the pearl intersects with a previous pearl, linking the pearl to the previous pearl, such that the ordered series of pearls forms a loop, and terminating growth of said ordered series of pearls.

12 The method of claim 1, wherein said step of generating an ordered series of pearls comprises:

• receiving an initial seed point and direction;
• and
• sequentially generating pearls starting at said initial seed point and moving in said direction along the tubular structure by iteratively estimating the center location and radius for each pearl to fit each pearl to the tubular structure, until a stop condition is met.

An apparatus for segmenting a tubular structure in a 2D image, comprising:

• means for generating an ordered series of pearls along the tubular structure, wherein each pearl is a 2D disk having a center location and radius determine based on local pixel intensities inside of the pearl.

A computer readable medium encoded with computer executable instructions for segmenting a tubular structure in a 2D image, the computer executable instructions defining a step comprising:

• generating an ordered series of pearls along the tubular structure, wherein each pearl is a 2D disk having a center location and radius determine based on local pixel intensities inside of the pearl.