Systems and methods for anatomic structure segmentation in image analysis
Systems and methods are disclosed for anatomic structure segmentation in image analysis, using a computer system. One method includes: receiving an annotation and a plurality of keypoints for an anatomic structure in one or more images; computing distances from the plurality of keypoints to a boundary of the anatomic structure; training a model, using data in the one or more images and the computed distances, for predicting a boundary in the anatomic structure in an image of a patient's anatomy; receiving the image of the patient's anatomy including the anatomic structure; estimating a segmentation boundary in the anatomic structure in the image of the patient's anatomy; and predicting, using the trained model, a boundary location in the anatomic structure in the image of the patient's anatomy by generating a regression of distances from keypoints in the anatomic structure in the image of the patient's anatomy to the estimated boundary.
1. A computer-implemented method of anatomic structure segmentation in image analysis, the method comprising:
receiving first image data of an anatomic structure of a patient, the first image data including one or more frames along a centerline formed by a plurality of pixels or voxels;
obtaining an estimation of a boundary of the anatomic structure;
transforming each of the one or more frames into a curvilinear planar representation (CPR) based on a geometry associated with the estimation;
determining one or more keypoints in each frame based on the estimation of the boundary; and
generating, using a trained model, a sub-pixel or sub-voxel boundary prediction of the anatomic structure in the first image data based on the first image data, the determined one or more keypoints, and the estimation of the boundary of the anatomic structure, wherein the trained model has been trained, based on (i) second image data of the anatomic structure of one or more individuals, (ii) one or more estimations of the boundary of the anatomic structure of the one or more individuals, and (iii) distances between the estimations of the boundary and a sub-pixel or sub-voxel location of the boundary of the anatomic structure of the one or more individuals, to learn an association between the estimation of the boundary and the sub-pixel or sub-voxel location of the boundary.
2. The method of claim 1 , wherein the association between the estimation of the boundary and the sub-pixel or sub-voxel location of the boundary is based on a regression of intensity values of pixels or voxels in the second image data along each of a plurality of radial angular directions out from one or more keypoints from the one or more estimations of the boundary.
3. The method of claim 2 , wherein the regression is a continuous regression.
4. The method of claim 2 , the regression on the intensity values includes:
performing the regression on a first column of columns of pixels or voxels in a respective segment of each frame in the second image data;
cyclically rotating the columns of pixels or voxels so that a last column is rearranged to be a replacement of the first column, and so that any remaining columns are shifted by one; and
repeating the performing of the regression and the cyclical rotation until the regression has been performed on each column of the frame.
5. The method of claim 1 , wherein the transforming of each frame into CPR is performed in segments.
6. The method of claim 1 , wherein the estimation of the boundary of the anatomic structure is in a format of a mesh, a voxel, an implicit surface representation, or a point cloud.
7. The method of claim 1 , wherein a relative location of the one or more keypoints is fixed relative to each frame.
8. The method of claim 1 , wherein:
the sub-pixel or sub-voxel boundary prediction includes a point cloud; and
the method further comprises generating a continuous surface of the boundary based on the point cloud, the continuous surface having sub-pixel or sub-voxel accuracy.
9. The method of claim 1 , wherein the anatomic structure comprises a blood vessel.
10. The method of claim 9 , wherein the boundary is associated with a vessel lumen boundary, a vessel lumen surface, or a combination thereof.
11. A system for anatomic structure segmentation in image analysis, comprising:
at least one processor; and
a memory operatively connected to the processor, and storing:
a trained model that has been trained, based on (i) first image data of an anatomic structure of one or more individuals, (ii) one or more estimations of a boundary of the anatomic structure of the one or more individuals, and (iii) distances between the estimation of the boundary and a sub-pixel or sub-voxel location of the boundary of the anatomic structure of the one or more individuals, to learn an association between the estimation of the boundary and the sub-pixel or sub-voxel location of the boundary; and
instructions executable by the processor to perform operations, including:
receiving second image data of an anatomic structure of a patient, the second image data including one or more frames along a centerline formed by a plurality of pixels or voxels;
obtaining an estimation of a boundary of the anatomic structure of the patient;
transforming each of the one or more frames into a curvilinear planar representation (CPR) based on a geometry associated with the estimation;
determining one or more keypoints in each frame based on the estimation; and
generating, using the trained model, a sub-pixel or sub-voxel boundary prediction of the anatomic structure in the second image data based on the second image data, the determined one or more keypoints, and the estimation of the boundary of the anatomic structure.
12. The system of claim 11 , wherein the association between the estimation of the boundary and the sub-pixel or sub-voxel location of the boundary is based on a regression of intensity values of pixels or voxels in the first image data along each of a plurality of radial angular directions out from one or more keypoints from the one or more estimations of the boundary.
13. The system of claim 12 , wherein the regression is a continuous regression.
14. The system of claim 11 , wherein:
the sub-pixel or sub-voxel boundary prediction includes a point cloud; and
the operations further include generating a continuous surface of the boundary based on the point cloud, the continuous surface having sub-pixel or sub-voxel accuracy.
15. A non-transitory computer readable medium for use on a computer system containing computer-executable programming instructions for anatomic structure segmentation in image analysis, the instructions executing operations comprising:
receiving first image data of an anatomic structure of a patient, the first image data including one or more frames along a centerline formed by a plurality of pixels or voxels;
obtaining an estimation of a boundary of the anatomic structure;
transforming each of the one or more frames into a curvilinear planar representation (CPR) based on a geometry associated with the estimation;
determining one or more keypoints in each frame based on the estimation of the boundary; and
generating, using a trained model, a sub-pixel or sub-voxel boundary prediction of the anatomic structure in the first image data based on the first image data, the determined one or more keypoints, and the estimation of the boundary of the anatomic structure, wherein the trained model has been trained, based on (i) second image data of the anatomic structure of one or more individuals, (ii) one or more estimations of the boundary of the anatomic structure of the one or more individuals, and (iii) distances between the estimations of the boundary and a sub-pixel or sub-voxel location of the boundary of the anatomic structure of the one or more individuals, to learn an association between the estimation of the boundary and the sub-pixel or sub-voxel location of the boundary.
16. The non-transitory computer readable medium of claim 15 , wherein the association between the estimation of the boundary and the sub-pixel or sub-voxel location of the boundary is based on a regression of intensity values of pixels or voxels in the second image data along each of a plurality of radial angular directions out from one or more keypoints from the one or more estimations of the boundary.
17. The non-transitory computer readable medium of claim 15 , wherein a relative location of the one or more keypoints is fixed relative to each frame.
18. The non-transitory computer readable medium of claim 15 , wherein:
the sub-pixel or sub-voxel boundary prediction includes a point cloud; and
the method further comprises generating a continuous surface of the boundary based on the point cloud, the continuous surface having sub-pixel or sub-voxel accuracy.
19. The non-transitory computer readable medium of claim 15 , wherein the anatomic structure comprises a blood vessel.
20. The non-transitory computer readable medium of claim 19 , wherein the boundary is associated with a vessel lumen boundary, a vessel lumen surface, or a combination thereof.