Image segmentation method

A method for efficient searching for optimal relative thresholds for use in identifying structures depicted in an image comprising voxels representing at least three different structure types A, B, and C, wherein a voxel representing structure type A is presumed to have a greater intensity associated therewith than an adjacent voxel representing structure type B, and wherein a voxel representing structure type B is presumed to have a greater intensity associated therewith than an adjacent voxel representing structure type C, the method comprising:

• specifying a plurality of candidate B-A criteria by which a comparison of the intensity associated with said first voxel with an intensity associated with a second voxel is to be considered to indicate that the first voxel represents structure type B assuming the second voxel represents structure type A;
• performing the following steps (a)-(d) with each of said candidate B-A criteria to obtain corresponding threshold-search voxels that are at least temporarily classified as representing structure type B:
• (a) selecting a first voxel to be classified as being either structure type A or B, said first voxel having an intensity associated therewith,
• (b) determining a B-A path of voxels passing through said first voxel,
• (c) defining a B-A maximum distance along said B-A path that is substantially less than the maximum distance allowed by the image, and
• (d) determining whether there is a second voxel on said B-A path, within said B-A maximum distance of said first voxel, such that the intensities associated with said first and second voxels together satisfy said B-A criteria, and if so, the method further comprises at least temporarily classifying said first voxel as representing structure type B;
• defining C-A criteria by which a comparison of the intensity associated with a third voxel with an intensity associated with a fourth voxel is to be considered to indicate that the third voxel represents structure type C assuming the fourth voxel represents structure type A;
• performing the following steps (e)-(i) for each of said C-A criteria:
• (e) determining a C-A path of voxels passing through said third voxel,
• (f) defining a C-A maximum distance along said C-A path that is substantially less than the maximum distance allowed by the image,
• (g) identifying a set of one or more of said threshold-search voxels, on said C-A path within said C-A maximum distance of said third voxel, the voxels of said set having respective intensities associated therewith,
• (h) determining a set of comparative values, each comparative value indicating a comparison of the intensity associated with a respective one of the voxels of said set of threshold-search voxels with the intensity associated with said fourth voxel, and
• (i) determining an average relative intensity that is, at least in part, based on the average of said comparative values;
• forming a plurality of lists of said threshold-search voxels, each list corresponding to a unique predetermined range of the average relative intensities obtained in steps (i);
• defining an objective function for the structure types A, B, and C;
• and
• re-classifying the voxels in said lists from structure type B to structure type C so as to lower the value of said objective function.