Method of Segmenting Anatomic Entities in 3D Digital Medical Images

• US 20100189320A1
• Filed: 06/10/2008
• Published: 07/29/2010
• Est. Priority Date: 06/19/2007
• Status: Active Grant

• Current Assignee: Agfa Nv (Agfa-Gevaert N.V.)
• Original Assignee: AGFA Healthcare NV (Agfa-Gevaert N.V.)
• Inventors: Dewaele, Piet
• Granted Patent: US 9,129,390 B2
• Time in Patent Office: Days
• US Class Current: 382/128
• CPC Class Codes: G06T 2207/10072 Tomographic images
• CPC Class Codes: G06T 2207/30004 Biomedical image processing
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/162 involving graph-based methods
• CPC Class Codes: G06T 7/168 involving transform domain methods

Abstract

For each of a number of landmarks in an image an initial position of the landmark is defined. Next a neighborhood around the initial position comprising a number of candidate locations of the landmark, is sampled and a cost is associated with each of the candidate locations. A cost function expressing a weighted sum of overall gray level cost and overall shape cost for all candidate locations is optimized. A segmented anatomic entity is defined as a path through a selected combination of candidate locations for which combination the cost function is optimized. During optimization towards the optimal segmented surface/volume graph traversal methods are exploited.

Claims

1. A method of segmenting an anatomic entity in a digital medical image comprising:

defining, for each of a number of landmarks in said image, an initial position of said landmark,

calculating a graph data structure for all landmark points, each node (vertex) representing a landmark and each edge (arc) representing a connection between a pair of landmarks;

calculating a path through the graph data structure either in a systematic way or in a random way;

sampling a neighborhood around said initial position, said neighborhood comprising a number of candidate locations of said landmark,

associating a cost with each of said candidate locations,

optimizing a cost function expressing a weighted sum of overall gray level cost and overall shape cost for all candidate locations,

calculating a segmented anatomic entity as a surface or volume through a selected combination of said candidate locations for each landmark for which combination said cost function is optimized, the order of visiting the landmarks determined by said path through the graph data structure, said path being a minimum spanning tree (MST).

2. A method according to claim 1 wherein said cost associated with a candidate location represents a measure of similarity between the sampled gray level pattern at said candidate location and an expected gray level pattern at said landmark, said similarity expressed by the Mahalanobis distance defined as the distance between the sampled gray level pattern and the mean gray level pattern weighted with the inverse covariance matrix.

3. A method according to claim 2 wherein said cost associated with a candidate location is the total gray level cost obtained by summing gray level costs of said candidate location in every feature image of a set of feature images calculated for said medical image, said gray level costs expressed as the Mahalanobis distance defined as the distance between the gray value profile at said candidate location with a corresponding mean profile weighted by the inverse of a corresponding covariance matrix, said mean profile and covariance matrix retrieved from a gray value model associated with said anatomic entity.

4. A method according to claim 3 wherein said gray value model is obtained by

sampling a manually segmented surface or volume of said anatomic entity at a number of landmark points;

forming a polygonal mesh of the surface or polyhedral mesh of the volume by connecting groups of neighboring landmark points in a polygon or respectively a polyhedron;

defining a graph data structure for all landmark points, each node (vertex) representing a landmark and each edge (arc) representing a connection between a pair of landmarks;

sampling a number of points in a neighborhood of each of said landmark points;

for each landmark point arranging sampled points in a neighborhood around a landmark point in a profile;

computing at least one feature image;

computing a mean profile for each landmark point and for each feature image;

computing a covariance matrix of the profiles for each landmark point and each feature image;

identifying said graph, mean profiles and covariance matrix as gray value model of said anatomic entity.

5. A method according to claim 4 where the polygon is a triangle.

6. A method according to claim 4 where the polyhedron is a tetrahedron.

7. A method according to claim 3 wherein a feature image is a derivative image at a predefined scale a or an image representation comprising mathematical moments of the local histogram of said image or derivative image in a window with width a around the location of a landmark.

8. A method according to claim 1 wherein said overall gray level cost is a combination of all individual gray level costs of a profile in a number of feature images whereby said individual cost is the Mahalanobis distance defined as the distance between said profile and the mean profile of that feature weighted with the inverse of the covariance matrix, said mean profile and covariance matrix being retrieved from a gray value model of said anatomic entity.

9. A method according to claim 8 wherein said gray value model is obtained by

sampling a manually segmented surface or volume of said anatomic entity at a number of landmark points;

forming a polygonal mesh of the surface or polyhedral mesh of the volume by connecting groups of neighboring landmark points in a polygon respectively a polyhedron;

defining a graph data structure for all landmark points, each node (vertex) representing a landmark and each edge (arc) representing a connection between a pair of landmarks;

sampling a number of points in a neighborhood of each of said landmark points;

for each landmark point arranging sampled points in a neighborhood around a landmark point in a profile;

computing at least one feature image;

computing a mean profile for each landmark point and for each feature image;

computing a covariance matrix of the profiles for each landmark point and each feature image;

identifying said graph, mean profiles and covariance matrix as gray value model of said anatomic entity.

10. A method according to claim 1 wherein said overall shape cost is a combination of all individual costs of connection vectors, said connection vectors connecting successive landmarks, for all landmarks whereby said individual shape cost is expressed by the Mahalanobis distance defined as the distance of a connection vector between two successive landmarks to the mean connection vector weighted with the inverse of the covariance matrix, said mean connection vector and covariance matrix being retrieved from a shape model of said anatomic entity.

11. A method according to claim 10 wherein said shape model is obtained by

sampling a manually segmented surface or volume of said anatomic entity at a number of landmark points;

forming a polygonal mesh of the surface or polyhedral mesh of the volume by connecting groups of neighboring landmark points in a polygon respectively a polyhedron;

defining a graph data structure for all landmark points, each node (vertex) representing a landmark and each edge (arc) representing a connection between a pair of landmarks;

computing connection vectors between pairs of landmark points;

computing a mean connection vector between pairs of landmark points;

computing a covariance matrix of connection vectors for each connection in the graph;

identifying said graph, mean connection vector and covariance matrices as a geometric model of said anatomic entity.

12. A method according to claim 1 wherein said cost function is optimized by dynamic programming.

13. A method according to claim 1 wherein said number of candidate locations is reduced by selecting those with minimal total gray value cost, said total gray value cost being the sum over all feature images of all gray value costs for the candidate position in said neighborhood, said gray value cost expressed as a Mahalanobis distance defined as the distance of a gray value profile at the candidate location to a corresponding mean profile weighted with the inverse of a covariance matrix, said mean profile and covariance matrix retrieved from a gray value model associated with said anatomic entity.

14. A method according to claim 1 wherein said neighborhood consists of a rectangular grid of sampling points.

15. A method according to claim 1, wherein said gray value model and said shape model are constructed from and applied to a multi-resolution representation of the image.

18. A method according to claim 1 amended in that the path through said graph data structure is obtained by breadth first traversal.

16. A method of measuring an anatomic entity in a digital medical image comprising

defining, for each of a number of landmarks in said image, an initial position of said landmark,

calculating a graph data structure for all landmark points, each node (vertex) representing a landmark and each edge (arc) representing a connection between a pair of landmarks;

calculating a path through the graph data structure either in a systematic way or in a random way;

sampling a neighborhood around said initial position, said neighborhood comprising a number of candidate locations of said landmark,

associating a cost with each of said candidate locations,

optimizing a cost function expressing a weighted sum of overall gray level cost and overall shape cost for all candidate locations,

calculating a segmented anatomic entity as a surface or volume through a selected combination of said candidate locations for each landmark for which combination said cost function is optimized, the order of visiting the landmarks determined by said path through the graph data structure, said path being a minimum spanning tree (MST);

calculating or selecting characteristic points based on the segmentation of said anatomic entity,

calculating measurement objects based on said characteristic points,

deriving measurements of said anatomic entity on the basis of said measurement objects.

19. A method according to claim 16 amended in that the path through said graph data structure is obtained by breadth first traversal.

17. A method of measuring an anatomic entity in a digital medical image comprising

defining, for each of a number of landmarks in said image, an initial position of said landmark,

calculating a graph data structure for all landmark points, each node (vertex) representing a landmark and each edge (arc) representing a connection between a pair of landmarks;

calculating a path through the graph data structure either in a systematic way or in a random way;

sampling a neighborhood around said initial position, said neighborhood comprising a number of candidate locations of said landmark,

associating a cost with each of said candidate locations,

optimizing a cost function expressing a weighted sum of overall gray level cost and overall shape cost for all candidate locations,

calculating a segmented anatomic entity as a surface or volume through a selected combination of said candidate locations for each landmark for which combination said cost function is optimized, the order of visiting the landmarks determined by said path through the graph data structure, said path being a minimum spanning tree (MST);

identifying said landmarks as measurement points,

calculating measurement objects based on said measurement points,

deriving measurements of said anatomic entity on the basis of said measurement objects.

20. A method according to claim 17 amended in that the path through said graph data structure is obtained by breadth first traversal.

21. A computer program product adapted to carry out a method comprising:

defining, for each of a number of landmarks in said image, an initial position of said landmark,

calculating a graph data structure for all landmark points, each node (vertex) representing a landmark and each edge (arc) representing a connection between a pair of landmarks;

calculating a path through the graph data structure either in a systematic way or in a random way;

sampling a neighborhood around said initial position, said neighborhood comprising a number of candidate locations of said landmark,

associating a cost with each of said candidate locations,

optimizing a cost function expressing a weighted sum of overall gray level cost and overall shape cost for all candidate locations,

calculating a segmented anatomic entity as a surface or volume through a selected combination of said candidate locations for each landmark for which combination said cost function is optimized, the order of visiting the landmarks determined by said path through the graph data structure, said path being a minimum spanning tree (MST).

22. A computer readable medium comprising computer executable program code adapted to carry out a method comprising:

defining, for each of a number of landmarks in said image, an initial position of said landmark,

calculating a graph data structure for all landmark points, each node (vertex) representing a landmark and each edge (arc) representing a connection between a pair of landmarks;

calculating a path through the graph data structure either in a systematic way or in a random way;

sampling a neighborhood around said initial position, said neighborhood comprising a number of candidate locations of said landmark,

associating a cost with each of said candidate locations,

optimizing a cost function expressing a weighted sum of overall gray level cost and overall shape cost for all candidate locations,

calculating a segmented anatomic entity as a surface or volume through a selected combination of said candidate locations for each landmark for which combination said cost function is optimized, the order of visiting the landmarks determined by said path through the graph data structure, said path being a minimum spanning tree (MST).