Method of compensation of respiratory motion in cardiac imaging

1. A computer implemented method of cardiac imaging, comprising the steps of:

a. registering a pre-operative 3D volumetric dataset of a subject's coronary arteries with an ECG-gated angiogram sequence of a subject's heart;

b. adjusting the registration of the pre-operative volumetric dataset with one of said ECG-gated angiograms using a respiratory motion reference model that optimizes 6 or fewer parameters that describe a rigid transform to estimate coronary pose parameters from similarity measures describing a distance between a set of 2D biplane fluoroscopic images and a coronary centerline in the pre-operative volumetric dataset;

c. registering the pre-operative 3D volumetric dataset of the subject's coronary arteries with additional angiogram sequences of the subject's heart by using the reference model to restrict optimization of image alignments; and

d. presenting registration results to respective cardiac imaging users,

wherein said reference model of respiratory motion is a patient-specific model trained in a preparatory training phase, including

acquiring an initial sequence of ECG-gated 2D biplane X-ray angiograms;

registering a pre-operative 3D segmentation of the coronary arteries with the ECG-gated 2D biplane X-ray angiogram sequence; and

estimating the pose of the vasculature in each sequence image by applying a feature-based rigid 2D/3D registration algorithm that best maps a pre-operative 3D centerline segmentation of the coronary arteries to the 2D biplane X-ray angiograms, wherein the coronary centerline is reconstructed from the pre-operative 3D segmentation,

wherein for each breathing phase, an estimate for the cardiac pose represented by a rigid transform that forms a linear sub-manifold embedded in a six dimensional space of a rigid 3D transformation is obtained from a basis of a d-dimensional linear sub-manifold, d<=6, identified from a first d eigenvectors corresponding to a d largest eigenvalues of a covariance matrix of the linear sub-manifold.

2. The computer implemented method of claim 1 , wherein the pre-operative 3D volumetric dataset comprises a 3D coronary tree image of the cardiac target site.

3. The computer implemented method of claim 2 , wherein the 3D coronary tree image is reconstructed from a pre-operative CT scan of the patient.

4. The computer implemented method of claim 1 , wherein registering a pre-operative 3D volumetric dataset of a subject's coronary arteries with an ECG-gated angiogram sequence of a subject's heart comprises overlaying the 3D coronary tree image onto the ECG-gated 2D fluoroscopic angiogram sequence of images and optimizing an alignment therebetween.

5. The computer implemented method of claim 4 , wherein

the respiratory motion model is defined using principle component analysis on cardiac imaging samples that are obtained using 2D/3D registration of a pre-operative 3D segmentation of the coronary arteries of the patient with an ECG-gated 2D fluoroscopic angiogram sequence.

6. A non-transitory program storage device readable by a computer, tangibly embodying a program of instructions executed by the computer to perform a method for cardiac imaging, the method comprising the steps of:

a. registering a pre-operative 3D volumetric dataset of a subject's coronary arteries with an ECG-gated angiogram sequence of a subject's heart;

b. adjusting the registration of the pre-operative volumetric dataset with one of said ECG-gated angiograms using a respiratory motion reference model that optimizes 6 or fewer parameters that describe a rigid transform to estimate coronary pose parameters from similarity measures describing a distance between a set of 2D biplane fluoroscopic images and a coronary centerline in the pre-operative volumetric dataset;

c. registering the pre-operative 3D volumetric dataset of the subject's coronary arteries with additional angiogram sequences of the subject's heart by using the reference model to restrict optimization of image alignments; and

d. presenting registration results to respective cardiac imaging users,

wherein said reference model of respiratory motion is a patient-specific model trained in a preparatory training phase, including

acquiring an initial sequence of ECG-gated 2D biplane X-ray angiograms;

registering a pre-operative 3D segmentation of the coronary arteries with the ECG-gated 2D biplane X-ray angiogram sequence; and

estimating the pose of the vasculature in each sequence image by applying a feature-based rigid 2D/3D registration algorithm that best maps a pre-operative 3D centerline segmentation of the coronary arteries to the 2D biplane X-ray angiograms, wherein the coronary centerline is reconstructed from the pre-operative 3D segmentation,

wherein for each breathing phase, an estimate for the cardiac pose represented by a rigid transform that forms a linear sub-manifold embedded in a six dimensional space of a rigid 3D transformation is obtained from a basis of a d-dimensional linear sub-manifold, d<=6, identified from a first d eigenvectors corresponding to a d largest eigenvalues of a covariance matrix of the linear sub-manifold.

7. The computer readable program storage device of claim 6 , wherein the pre-operative 3D volumetric dataset comprises a 3D coronary tree image of the cardiac target site.

8. The computer readable program storage device of claim 7 , wherein the 3D coronary tree image is reconstructed from a pre-operative CT scan of the patient.

9. The computer readable program storage device of claim 6 , wherein registering a pre-operative 3D volumetric dataset of a subject's coronary arteries with an ECG-gated angiogram sequence of a subject's heart comprises overlaying the 3D coronary tree image onto the ECG-gated 2D fluoroscopic angiogram sequence of images and optimizing an alignment therebetween.

10. The computer readable program storage device of claim 9 , wherein the respiratory motion model is defined using principle component analysis on cardiac imaging samples that are obtained using 2D/3D registration of a pre-operative 3D segmentation of the coronary arteries of the patient with an ECG-gated 2D fluoroscopic angiogram sequence.

11. A method for cardiac imaging, comprising the steps of:

registering a pre-operative 3D volumetric dataset of a subject's coronary arteries with an ECG-gated angiogram sequence of a subject's heart;

adjusting the registration of the pre-operative volumetric dataset with one of said ECG-gated angiograms using a respiratory motion reference model; and

registering the pre-operative 3D volumetric dataset of the subject's coronary arteries with additional angiogram sequences of the subject's heart by using the reference model to restrict optimization of image alignments; and presenting registration results to respective cardiac imaging users,

wherein said reference model of respiratory motion is a patient-specific model trained by acquiring an initial sequence of ECG-gated 2D biplane X-ray angiograms;

registering a pre-operative 3D segmentation of the coronary arteries with the ECG-gated 2D biplane X-ray angiogram sequence; and

estimating the pose of the vasculature in each sequence image by applying a feature-based rigid 2D/3D registration algorithm that best maps a pre-operative 3D centerline segmentation of the coronary arteries to the 2D biplane X-ray angiograms,

wherein the coronary centerline is reconstructed from the pre-operative 3D segmentation,

wherein for each breathing phase, an estimate for the cardiac pose represented by a rigid transform that forms a linear sub-manifold embedded in a six dimensional space of a rigid 3D transformation is obtained from a basis of a d-dimensional linear sub-manifold, d<=6, identified from a first d eigenvectors corresponding to a d largest eigenvalues of a covariance matrix of the linear sub-manifold.

12. The method of claim 11 , wherein the respiratory motion reference model optimizes 6 or fewer parameters that describe a rigid transform to estimate coronary pose parameters from similarity measures describing a distance between a set of 2D biplane fluoroscopic images and a coronary centerline in the pre-operative volumetric dataset.