Combined cardiac and respiratory motion compensation for atrial fibrillation ablation procedures
A method for compensating cardiac and respiratory motion in atrial fibrillation ablation procedures includes (a) simultaneously determining a position of a circumferential mapping (CFM) catheter and a coronary sinus (CS) catheter in two consecutive image frames of a series of first 2-D image frames; (b) determining a distance between a virtual electrode on the CS catheter and a center of the CFM catheter for a first image frame of the two consecutive image frames, and for a second image frame of the two consecutive image frames; and (c) if an absolute difference of the distance for the first image frame and the distance for the second image frame is greater than a predetermined threshold, compensating for motion of the CFM catheter in a second 2-D image.
1. A computer implemented method for compensating cardiac and respiratory motion in atrial fibrillation ablation procedures, the method executed by a computer comprising the step of:
(a) acquiring a series of first 2-D image frames using an imaging system;
(b) simultaneously determining a position of a circumferential mapping (CFM) catheter and a coronary sinus (CS) catheter in two consecutive frames of the series of first 2-D image frames;
(c) determining a distance between a virtual electrode on the CS catheter and a center of the CFM catheter for a first consecutive image frame of the two consecutive image frames, and for a second consecutive image frame of the two consecutive image frames; and
(d) if an absolute difference of the distance for the first consecutive image frame and the distance for the second consecutive image frame is greater than a predetermined threshold, compensating for motion of the CFM catheter in another 2-D image,
wherein determining a position of the coronary sinus (CS) catheter comprises:
receiving a model of the CS catheter determined from a set of points extracted from a first image frame of the series of first 2-D image frames, wherein said CS catheter is modeled as a set of electrodes extending from a tip of the CS catheter to a first proximal electrode (FPE), and including a virtual electrode (VE) positioned at a first proximal location on the CS catheter;
detecting the tip, electrodes, and body points of the CS catheter in each subsequent frame using trained classifiers;
generating a set of hypothesis v t for locating the VE in each subsequent frame based on a location of the detected FPE; and
determining a hypothesis {circumflex over (v)} t for locating the VE in each subsequent frame by maximizing a posterior probability P(v t |I 0, . . . , t ) of the hypotheses v t for the image frames I 0, . . . , t .
2. The method of claim 1 , further comprising repeating steps (a), (b), (c), and (d) for each pair of consecutive image frames of the series of first 2-D image frames.
3. The method of claim 1 , wherein determining a position of the circumferential mapping (CFM) catheter comprises:
receiving a model of the CFM catheter determined from a set of points extracted from a first image frame of the series of first 2-D image frames, wherein said set of points are connected as a B-spline;
segmenting the CFM catheter in a region of interest in each subsequent image frame of the series of first 2-D image frames;
processing a segmentation result for each subsequent image frame to calculate a distance map for each subsequent image frame that encodes for each pixel in the image frame a distance from that pixel to a nearest pixel of the segmented CFM catheter; and
registering said CFM catheter model to said segmented CFM catheter in each subsequent image frame to determine an optimal translation vector between the CFM catheter and the segmented CFM catheter for each subsequent image frame.
4. The method of claim 3 , wherein a position of the region of interest in each subsequent image frame is determined from a center of the registered CFM catheter model in a previous image frame.
5. The method of claim 3 , wherein processing a segmentation result for each subsequent image frame comprises skelotonizing the segmentation result for each subsequent image frame and calculating the distance map of each subsequent image frame from the skelotonized segmentation result.
6. The method of claim 3 , wherein registering said CFM catheter model to said segmented CFM catheter in each subsequent image frame comprises minimizing a cost function
∑
i
I
DT
,
t
(
q
i
+
r
t
)
of the distance map to determine the optimal translation vector {circumflex over (r)} t , wherein I DT,t is a distance map of a subsequent image t, r t is a pixel in said distance map, q i is a point on the CFM catheter model, and the sum is over all points on the CFM catheter model.
7. The method of claim 1 , further comprising providing the set of detected electrodes and tips at each subsequent frame to a non-maximal suppression stage to reduce a number of catheter electrodes or tip candidates prior to selecting a catheter electrode or tip candidates having highest scores.
8. The method of claim 1 , wherein generating a set of hypotheses v t for locating the VE in each subsequent frame comprises:
generating a seed hypothesis v t for localizing the VE for a current time step tεN by translating a VE position from a previous frame, {circumflex over (v)} t-1 , to a position of the first proximal electrode, m FPE for the current time step; and
generating a set of tracking hypotheses v t from an affine transformation
v
t
=
(
R
t
0
T
1
)
·
v
_
t
,
with RεR 2×2 , tεR 2 being the parameter of the affine transformation, and wherein m FPE is a transformation center of the affine transformation.
9. The method of claim 1 , wherein said posterior probability P(v t |I 0, . . . , t ) for an image frame t is a product of a likelihood term P(I t |v t ), a prediction term, P(v t |v t-1 ), and a posterior probability P(v t-1 |I 0, . . . , t ) for a previous image frame t−1, wherein said likelihood term P(I t |v t ) is estimated by combining a VE detection probability and a VE location matching.
10. A computer implemented method for compensating cardiac and respiratory motion in atrial fibrillation ablation procedures, the method executed by a computer comprising the steps of:
(a) acquiring a series of first 2-D image frames using an imaging system;
(b) simultaneously determining a position of a circumferential mapping (CFM) catheter and a coronary sinus (CS) catheter in two consecutive image frames of a series of first 2-D image frames; wherein
determining a position of the circumferential mapping (CFM) catheter comprises registering a CFM catheter model to a segmented CFM catheter in each image frame to determine an optimal translation vector between the CFM catheter and the segmented CFM catheter for each image frame; and
determining a position of the coronary sinus (CS) catheter comprises determining a hypothesis {circumflex over (v)} t for locating a virtual electrode (VE) positioned at a first proximal location on the CS catheter in each frame by maximizing a posterior probability P(v t |I 0, . . . , t ) of a set of tracking hypotheses v t for the VE for the image frames I 0, . . . , t .
11. The method of claim 10 , further comprising:
(c) determining a distance between a virtual electrode on the CS catheter and a center of the CFM catheter for a first consecutive image frame of the two consecutive image frames, and for a second consecutive image frame of the two consecutive image frames; and
(d) if an absolute difference of the distance for the first consecutive image frame and the distance for the second consecutive image frame is greater than a predetermined threshold, compensating for motion of the CFM catheter in another 2-D image,
wherein steps (a), (b), (c), and (d) are repeated for each pair of consecutive image frames of the series of first 2-D image frames.
12. The method of claim 10 , wherein determining a position of the circumferential mapping (CFM) catheter further comprises:
receiving a model of the CFM catheter determined from a set of points extracted from a first image frame of the series of first 2-D image frames, wherein said set of points are connected as a B-spline;
segmenting the CFM catheter in a region of interest in each subsequent image frame of the series of first 2-D image frames; and
processing a segmentation result for each subsequent image frame to calculate a distance map for each subsequent image frame that encodes for each pixel in the image frame a distance from that pixel to a nearest pixel of the segmented CFM catheter,
wherein registering said CFM catheter model to said segmented CFM catheter in each image frame comprises minimizing a cost function
∑
i
I
DT
,
t
(
q
i
+
r
t
)
of the distance map to determine the optimal translation vector {circumflex over (r)} t , wherein I DT,t is a distance map of an image t, r t is a pixel in said distance map, q i is a point on the CFM catheter model, and the sum is over all points on the CFM catheter model.
13. The method of claim 10 , wherein determining a position of the coronary sinus (CS) catheter further comprises:
receiving a model of the CS catheter determined from a set of points extracted from a first image frame of the series of first 2-D image frames, wherein said CS catheter is modeled as a set of electrodes extending from a tip of the CS catheter to a first proximal electrode (FPE), and includes the virtual electrode (VE);
detecting the tip, electrodes, and body points of the CS catheter in each subsequent frame using trained classifiers; and
generating the set of tracking hypotheses v t for locating the VE in each subsequent frame based on a location of the detected FPE.
14. The method of claim 13 , wherein generating a set of tracking hypotheses v t for locating the VE in each subsequent frame comprises:
generating a seed hypothesis v t for localizing the VE for a current time step tεN by translating a VE position from a previous frame, {circumflex over (v)} t-1 , to a position of the first proximal electrode, m FPE for the current time step; and
generating a set of tracking hypotheses v t from an affine transformation
v
t
=
(
R
t
0
T
1
)
·
v
_
t
,
with RεR 2×2 , tεR 2 being the parameter of the affine transformation, and wherein m FPE is a transformation center of the affine transformation.
15. A non-transitory program storage device readable by a computer, tangibly embodying a program of instructions executed by the computer to perform the method steps for compensating cardiac and respiratory motion in atrial fibrillation ablation procedures, the method comprising the steps of:
(a) acquiring a series of first 2-D image frames using an imaging system;
(b) simultaneously determining a position of a circumferential mapping (CFM) catheter and a coronary sinus (CS) catheter in two consecutive image frames of a series of first 2-D image frames; wherein
determining a position of the circumferential mapping (CFM) catheter comprises registering a CFM catheter model to a segmented CFM catheter in each image frame to determine an optimal translation vector between the CFM catheter and the segmented CFM catheter for each image frame; and
determining a position of the coronary sinus (CS) catheter comprises determining a hypothesis {circumflex over (v)} t for locating a virtual electrode (VE) positioned at a first proximal location on the CS catheter in each frame by maximizing a posterior probability P(v t |I 0, . . . , t ) of a set of tracking hypotheses v t for the VE for the image frames I 0, . . . , t .
16. The computer readable program storage device of claim 15 , the method further comprising:
(c) determining a distance between a virtual electrode on the CS catheter and a center of the CFM catheter for a first consecutive image frame of the two consecutive image frames, and for a consecutive second image frame of the two consecutive image frames;
(d) if an absolute difference of the distance for the first consecutive image frame and the distance for the second consecutive image frame is greater than a predetermined threshold, compensating for motion of the CFM catheter in another 2-D image; and
repeating steps (a), (b), (c), and (d) for each pair of consecutive image frames of the series of first 2-D image frames.
17. The computer readable program storage device of claim 15 , wherein determining a position of the circumferential mapping (CFM) catheter comprises:
receiving a model of the CFM catheter determined from a set of points extracted from a first image frame of the series of first 2-D image frames, wherein said set of points are connected as a B-spline;
segmenting the CFM catheter in a region of interest in each subsequent image frame of the series of first 2-D image frames; and
processing a segmentation result for each subsequent image frame to calculate a distance map for each subsequent image frame that encodes for each pixel in the image frame a distance from that pixel to a nearest pixel of the segmented CFM catheter.
18. The computer readable program storage device of claim 17 , wherein a position of the region of interest in each subsequent image frame is determined from a center of the registered CFM catheter model in a previous image frame.
19. The computer readable program storage device of claim 17 , wherein processing a segmentation result for each subsequent image frame comprises skeletonizing the segmentation result for each subsequent image frame and calculating the distance map of each subsequent image frame from the skeletonized segmentation result.
20. The computer readable program storage device of claim 17 , wherein registering said CFM catheter model to said segmented CFM catheter in each subsequent image frame comprises minimizing a cost function
∑
i
I
DT
,
t
(
q
i
+
r
t
)
of the distance map to determine the optimal translation vector {circumflex over (r)} t wherein I DT,t is a distance map of a subsequent image t, r t is a pixel in said distance map, q i is a point on the CFM catheter model, and the sum is over all points on the CFM catheter model.
21. The computer readable program storage device of claim 15 , wherein determining a position of the coronary sinus (CS) catheter comprises:
receiving a model of the CS catheter determined from a set of points extracted from a first image frame of the series of first 2-D image frames, wherein said CS catheter is modeled as a set of electrodes extending from a tip of the CS catheter to a first proximal electrode (FPE), and including the virtual electrode (VE) positioned at a most proximal location on the CS catheter;
detecting the tip, electrodes, and body points of the CS catheter in each subsequent frame using trained classifiers; and
generating a set of hypotheses v t for locating the VE in each subsequent frame based on a location of the detected FPE.
22. The computer readable program storage device of claim 21 , the method further comprising providing the set of detected electrodes and tips at each subsequent frame to a non-maximal suppression stage to reduce a number of catheter electrodes or tip candidates prior to selecting a catheter electrode or tip candidates having highest scores.
23. The computer readable program storage device of claim 21 , wherein generating a set of hypotheses v t for locating the VE in each subsequent frame comprises:
generating a seed hypothesis v t for localizing the VE for a current time step tεN by translating a VE position from a previous frame, {circumflex over (v)} t-1 , to a position of the first proximal electrode, m FPE for the current time step; and
generating a set of tracking hypotheses v t from an affine transformation
v
t
=
(
R
t
0
T
1
)
·
v
_
t
,
with RεR 2×2 , tεR 2 being the parameter of the affine transformation, and wherein m FPE is a transformation center of the affine transformation.
24. The computer readable program storage device of claim 21 , wherein said posterior probability P(v t |I 0, . . . , t ) for an image frame t is a product of a likelihood term P(I t |v t ), a prediction term, P(v t |v t-1 ), and a posterior probability P(v t-1 |I 0, . . . , t ) for a previous image frame t−1, wherein said likelihood term P(I t |v t ) is estimated by combining a VE detection probability and a VE location matching.