Tree structure based 2D to 3D registration
First Claim
1. A method for tree structure based two-dimensional to three-dimensional registration comprising:
- receiving three-dimensional (3D) image data indicative of vessels;
segmenting the 3D image data with a vessel tree to generate a segmented 3D image;
simulating two-dimensional (2D) image from the segmented 3D image data and a hypothesized pose to form a simulated 2D image;
receiving 2D data indicative of vessels;
segmenting the received 2D image data with a vessel tree to generate a segmented 2D image;
extracting a 2D distance map image from a distance transform of the segmented 2D image;
comparing the simulated 2D image with the 2D distance map image to generate a probability of the hypothesized pose;
optimizing the hypothesized pose based on a particle filtering scheme that uses a plurality of poses of the received 2D image data relative to the received 3D image data and the probability; and
registering the received 2D image data with the received 3D image data based on the optimized hypothesized pose,wherein weights of the particle filtering scheme are based on the distance transform of the segmented 2D data, andwherein the method is performed by a processor.
2 Assignments
0 Petitions
Accused Products
Abstract
A system and method for tree structure based two-dimensional to three-dimensional registration are provided for receiving three-dimensional (3D) data indicative of vessels, segmenting the 3D data with a vessel tree, simulating two-dimensional (2D) data responsive to the segmented 3D data to form a simulated 2D image, receiving 2D data indicative of vessels, segmenting the received 2D data with a vessel tree, finding a distance transform of the segmented 2D data to form a 2D distance map image, considering a set of poses as state vectors with the corresponding probability computed using the similarity measure computation, re-sampling the pose vector to find a set of most probable poses and considering them as hypothesized poses, recomputing the projection and re-evaluating the probability of the hypothesized poses and updating the state vector until the optimal pose is found.
-
Citations
30 Claims
-
1. A method for tree structure based two-dimensional to three-dimensional registration comprising:
-
receiving three-dimensional (3D) image data indicative of vessels; segmenting the 3D image data with a vessel tree to generate a segmented 3D image; simulating two-dimensional (2D) image from the segmented 3D image data and a hypothesized pose to form a simulated 2D image; receiving 2D data indicative of vessels; segmenting the received 2D image data with a vessel tree to generate a segmented 2D image; extracting a 2D distance map image from a distance transform of the segmented 2D image; comparing the simulated 2D image with the 2D distance map image to generate a probability of the hypothesized pose; optimizing the hypothesized pose based on a particle filtering scheme that uses a plurality of poses of the received 2D image data relative to the received 3D image data and the probability; and registering the received 2D image data with the received 3D image data based on the optimized hypothesized pose, wherein weights of the particle filtering scheme are based on the distance transform of the segmented 2D data, and wherein the method is performed by a processor. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21)
-
-
22. A method for two-dimensional to three-dimensional registration, the method comprising:
-
receiving three-dimensional (3D) data indicative of vessels; segmenting the 3D data with a vessel tree; simulating two-dimensional (2D) data responsive to the segmented 3D data to form a simulated 2D image by the computer; receiving 2D data indicative of vessels; segmenting the received 2D data with a vessel tree; finding a distance transform of the segmented 2D data to form a 2D distance map image; using a similarity metric based on the distance map to compare the simulated 2D projection image with the segmented 2D image; computing a probability based on the similarity metric, and re-sampling poses of the received 2D data relative to the received 3D data based on a particle filtering scheme to keep the most probable poses generating the highest similarity between the simulated 2D image and segmented 2D image; and finding the pose with the highest similarity for registering the received 2D data to the corresponding projection of the received 3D data, wherein the particle-filtering comprises; using an initial particle distribution as a prior vector of pose parameters and probability and considering these particles as current state particles; adding noise to the state particles to dislocate them randomly and computing new weights associated with the modified pose particles, where the weight computation is based on a similarity measure computed from the distance transform of the 2D segmented projection image that is only computed once at the beginning of the operation, the weights indicating the probability of a particle'"'"'s pose parameters being accurate; and processing the current state and the modified particles into a re-sampling step to normalize the weights, while considering these particles as the new current state particles, wherein the method is performed by a processor.
-
-
23. A system for tree structure based two-dimensional to three-dimensional registration comprising:
-
a processor; a segmentation software module for segmenting 2D data and 3D data with respective vessel trees; and a distance transform computation software module for computing a distance transform of the segmented 2D data to form a 2D distance map image; a similarity measure computation software module to generate a similarity metric based on the 2D distance map image and a projection of the 3D data; and an optimization module, based on a particle filtering scheme with pose parameters as a state vector to hypothesize a set of poses, wherein the optimization module is configured to re-sample poses of the pose parameters based on the similarity metric to find an optimal pose that registers the received 2D data relative to the projection of the 3D data, wherein weights of the particle filtering scheme are based on the distance transform of the segmented 2D data. - View Dependent Claims (24)
-
-
25. A program storage device readable by machine, tangibly embodying a program of instructions executable by the machine to perform program steps for tree structure based two-dimensional to three-dimensional registration, the program steps comprising:
-
receiving three-dimensional (3D) data indicative of vessels; segmenting the 3D data with a vessel tree; simulating two-dimensional (2D) data responsive to the segmented 3D data to form a simulated 2D image; receiving 2D data indicative of vessels; segmenting the received 2D data with a vessel tree; finding a distance transform of the segmented 2D data to form a 2D distance map image; computing a similarity measure based on the simulated 2D image and the 2D distance map image; optimizing pose parameters based on a particle filtering scheme using the pose parameters as state vectors and by evaluating a probability based the similarity measure, and re-sampling the state vectors to register the simulated 2D image with the segmented 2D data, wherein weights of the particle filtering scheme are based on the distance transform of the segmented 2D data. - View Dependent Claims (26, 27, 28, 29, 30)
-
Specification