Medical image segmentation
A segmentation method comprises clustering spatial, intensity and volumetric shape index to automatically segment a medical lesion. The algorithm has the following steps: (1) calculating volumetric shape index (SI) for each voxel in the image; (2) combining the SI features with the intensity range and the spatial position (x, y, z) to form a 5-dimensional feature vector set; (3) grouping the 5-dimensional feature vector set into clusters; (4) employing a modified expectation-maximization algorithm (EM) considering not only spatial but also shape features on an intensity mode map from the clustering algorithm to merge the neighbouring regions or modes. The joint spatial-intensity-shape feature provides rich information for the segmentation of the anatomic structures of interest, such as lesions or tumours.
1. A method for segmenting a three-dimensional medical image, comprising:
receiving, by a computing device, medical image data representing spatial variation of the three-dimensional medical image;
deriving, by the computing device, shape data from the medical image data;
clustering, by the computing device, image data including the medical image data and the shape data, to generate a plurality of modes;
merging neighboring ones of said plurality of modes to form merged modes; and
outputting, by the computing device, segmentation data relating to the three-dimensional medical image, derived from the plurality of modes, wherein the segmentation data is based on the merged modes.
2. The method of claim 1 , wherein the medical image data represents a spatial intensity variation of the three-dimensional medical image.
3. The method of claim 1 , wherein the shape data represents a volumetric shape index of the three-dimensional medical image.
4. The method of claim 1 , wherein the shape data is based on a Gaussian curvature.
5. The method of claim 1 , wherein the shape data is based on a mean curvature.
6. The method of claim 1 , wherein the shape data is determined from a Hessian matrix.
7. The method of claim 6 , wherein the shape data is calculated from eigenvalues of the Hessian matrix.
8. The method of claim 1 , wherein said clustering comprises clustering a set of vectors each comprising a spatial location within the three-dimensional image, a corresponding medical image datum, and a corresponding shape datum.
9. The method of claim 1 , wherein said clustering generates a mode map corresponding to the medical image data.
10. The method of claim 1 , wherein said clustering generates a mode map corresponding to the shape data.
11. The method of claim 1 , wherein said clustering comprises a mean shift clustering step.
12. The method of claim 1 , wherein said clustering comprises a subtractive clustering step.
13. The method of claim 1 , wherein said clustering uses an adaptive kernel window size.
14. The method of claim 1 , wherein said merging comprises an expectation-maximization step.
15. The method of claim 14 , wherein the expectation, maximization step is based on the shape data.
16. The method of claim 14 , wherein the expectation-maximization step includes calculating a spatial prior probability.
17. The method of claim 16 , wherein the spatial prior probability involves a spatial constraint.
18. The method of claim 17 , wherein the spatial constraint is imposed by a Markov Random Field.
19. The method of claim 17 , wherein the spatial constraint is imposed by a Gibbs random field.
20. The method of claim 16 , wherein the spatial prior probability involves a shape constraint.
21. The method of claim 1 , wherein said merging comprises a graph- theoretic step.
22. The method of claim 1 , wherein at least one of said receiving deriving, clustering; and outputting is performed automatically by the computing device.
23. The method of claim 1 , wherein the three-dimensional medical image is an image of at least part of a colon.
24. The method of claim 23 , wherein the segmentation data extracts a polyp from the three-dimensional medical image.
25. The method of claim 1 , wherein the three-dimensional medical image is an image of at least part of a lung.
26. The method of claim 25 , wherein the segmentation data extracts a lesion from the three-dimensional medical image.
27. The method of claim 1 , wherein the three-dimensional medical image is a computed tomography (CT) image.
28. The method of claim 1 , wherein the three-dimensional medical image is a magnetic resonance (MR) image.
29. The method of claim 1 , wherein the three-dimensional medical image is an X-ray image.
30. The method of claim 1 , wherein the three-dimensional medical image is an ultrasound image.
31. A method for segmenting a three-dimensional medical image, comprising:
receiving, by a computing device, medical image data representing a spatial variation of a property of the three-dimensional medical image;
clustering, by the computing device, the medical image data to generate a plurality of modes;
merging, by the computing device, neighbouring ones of the plurality of modes to form merged modes; and
outputting, by the computing device, segmentation data relating to the three-dimensional medical image, derived from the merged modes.
32. A method for segmenting a three-dimensional medical image, comprising:
receiving, by a computing device, medical image data representing the intensity of a plurality of voxels of the three-dimensional medical image;
deriving, by the computing device, shape data for respective voxels from the medical image data;
composing, by the computing device, a plurality of vectors corresponding to the voxels, each vector comprising three-dimensional coordinates, intensity data, and the-shape data for a corresponding voxel;
clustering, by the computing device, the plurality of vectors to generate a plurality of modes;
merging, by the computing device, neighbouring ones of the plurality of modes to form merged modes; and
outputting, by the computing device, segmentation data relating to the three-dimensional medical image, derived from the merged modes.
33. A method for segmenting a three-dimension medical image, comprising;
clustering, by a computing device, spatial feature data of the three-dimensional medical image, including shape data,
merging neighboring results of the clustering, and
segmenting, by the computing device, the three-dimensional medical image based on said clustering and said merging.
34. A medical imaging apparatus, comprising:
a memory and processor configured to:
receive medical image data representing spatial variation of the three-dimensional medical image;
derive shape data from the medical image data;
cluster image data, including the medical image and the shape data, to generate a plurality of modes;
merge neighboring ones of said plurality of modes to form merged modes; and
output segmentation data relating to the three-dimensional medical image, derived from the plurality of modes, wherein the segmentation data is based on the merged modes.
35. A non-transitory computer-readable storage medium having stored thereon computer executable instructions, execution of which by a computing device causes the computing device to perform operations for segmenting a three-dimensional medical image, the operations comprising:
receiving medical image data representing spatial variation of the three-dimensional medical image;
deriving shape date from the medical image data;
clustering image data, including the medical image data and the shape data, to generate a plurality of modes;
merging neighboring ones of said plurality of modes to form merged modes; and
outputting segmentation data relating to the three-dimensional medical image, derived from the plurality of modes, wherein the segmentation data is based on the merged modes.
36. A medical imaging apparatus, comprising:
means for receiving medical image data representing spatial variation of the three-dimensional medical image;
means for deriving shape data from the medical image data;
means for clustering image data, including the medical image data and the shape data, to generate a plurality of modes;
means for merging neighboring ones of said plurality of modes to form merged modes; and
means for outputting segmentation data relating to the three-dimensional medical image, derived from the plurality of modes, wherein the segmentation data is based on the merged modes.