System and method for online optimization of guidewire visibility in fluoroscopic systems

1. A method for online optimization of guidewire visibility in fluoroscopic images, comprising the steps of:

providing an digitized image acquired from a fluoroscopic imaging system, said image comprising an array of intensities corresponding to a 2-dimensional grid of pixels;

detecting a guidewire in said fluoroscopic image;

enhancing the visibility of the guidewire in said fluoroscopic image;

calculating a visibility measure of said guidewire in said fluoroscopic image; and

readjusting acquisition parameters of said fluoroscopic imaging system wherein said guidewire visibility is improved,

wherein enhancing guidewire visibility comprises constructing a mask image I M wherein pixels that have high filter response and are at most one pixel from said guidewire are also given value 1 and value 0 elsewhere, blurring said mask I M over every 3×3 pixel neighborhood, and multiplying the blurred mask by 0.1×(max(I)−min(I)) and subtracting the result from said fluoroscopic image.

2. The method of claim 1 , wherein detecting a guidewire comprises filtering said image with one or more steerable filters of different angular orientations, assigning a weight W and orientation O for each pixel based on said filter response for each pixel, wherein each pixel weight is assigned to a function of a maximum filter response magnitude and said pixel orientation is calculated from the angle producing said maximum filter response if said magnitude is greater than zero, wherein guidewire pixels have a higher weight than non-guidewire pixels.

3. The method of claim 1 , wherein blurring said mask over a 3×3 pixel neighborhood comprises multiplying 3×3 sub-matrices of said mask by the following 3×3 matrix:

[

0.0625

0.1250

0.0625

0.1250

0.2500

0.1250

0.0625

0.1250

0.0625

]

4. The method of claim 1 , wherein calculating a visibility measure comprises:

selecting a set of pixels on said guidewire;

selecting a set of background pixels in a neighborhood of said guidewire that do not include the guidewire pixels;

calculating a mean intensity μ N , μ B , and intensity variance σ N , σ B for both the guidewire pixels and the background pixels, respectively;

calculating a distance measure d 2 between said 2 sets of pixels as

d

2

⁡

(

w

)

=

1

4

⁢

(

σ

N

-

σ

B

)

2

⁢

(

σ

N

+

σ

B

)

-

1

+

1

2

⁢

log

⁢

1

2

⁢

(

σ

N

+

σ

B

)

σ

N

⁢

σ

B

,

wherein w represents the guidewire pixels; and

calculating a visibility V of said guidewire from

V

=

1

n

⁢

∑

i

=

1

n

⁢

⁢

d

2

⁡

(

w

i

)

,

wherein the sum is over the guidewire pixels.

5. The method of claim 4 , further comprising adding a contrast term ε to said distance measure:

d

2

⁡

(

w

)

=

1

4

⁢

(

μ

N

-

μ

B

)

2

⁢

(

σ

N

+

σ

B

+

ɛ

)

-

1

+

1

2

⁢

log

⁢

1

2

⁢

(

σ

N

+

σ

B

)

σ

N

⁢

σ

B

.

wherein a relative importance of the two terms is tuned by a value of said ε-term.

6. A computer-implemented method for online optimization of guidewire visibility in fluoroscopic images, the method performed by a computer comprising the steps of:

providing an digitized image acquired from a fluoroscopic imaging system, said image comprising an array of intensities corresponding to a 2-dimensional grid of pixels;

detecting a guidewire in said fluoroscopic image;

computing a pixel value vector for the guidewire for a sequence of windows centered about and normal to the guidewire, averaged over a wire width;

computing pixel value vectors for backgrounds b 0 , b 1 on either side of the guidewire for said sequence of windows, averaged over background widths;

calculating mean vectors μ N , μ b 0 , μ b 1 as a moving average of the moving window, and calculating variances σ N , σ B from the foreground and background means; and

calculating a distance between said guidewire pixels and said background pixels from

d

2

⁡

(

w

)

=

1

8

⁢

⁢

n

⁢

∑

i

=

1

n

⁢

⁢

[

(

μ

N

i

-

μ

b

⁢

⁢

0

i

)

2

+

(

μ

N

i

-

μ

b

⁢

⁢

1

i

)

2

]

⁢

(

σ

N

+

σ

B

+

ɛ

)

-

1

+

1

2

⁢

log

⁢

1

2

⁢

(

σ

N

+

σ

B

)

σ

N

⁢

σ

B

wherein w represents the guidewire pixels, the sum is over the sequence of windows, and n is the number of windows; and

calculating a visibility V of said guidewire from

V

=

1

n

⁢

∑

i

=

1

n

⁢

⁢

d

2

⁡

(

w

i

)

,

wherein the sum is over the guidewire pixels.

7. The method of claim 6 , further comprising readjusting acquisition parameters of said fluoroscopic imaging system wherein said guidewire visibility is improved.

8. The method of claim 6 , further comprising normalizing the fluroscopic image to take on values in the range from 0 to 1.

9. The method of claim 6 , further comprising blurring said image with a Gaussian to remove high frequency noise.

10. The method of claim 6 , further comprising constructing a mask image I M wherein pixels that have high filter response and are close to said guidewire are also given value 1 and value 0 elsewhere, blurring said mask I M over every 3×3 pixel neighborhood, and multiplying the blurred mask by 0.1×(max(I)−min(I)) and subtracting the result from said fluoroscopic image, wherein the visibility of the guidewire in said fluoroscopic image is enhanced.

11. The method of claim 6 , further comprising adding a contrast term ε to the (σ N +σ B ) −1 factor of distance measure, wherein a relative importance of the two terms is tuned by a value of said ε-term.

12. The method of claim 6 , wherein calculating a distance measure between said 2 sets of pixels comprises:

normalizing the fluroscopic image to take on values in the range from 0 to 1;

blurring said image with a Gaussian to remove high frequency noise;

computing a pixel value vector for the guidewire for a sequence of windows centered about and normal to the guidewire, averaged over a wire width;

computing pixel value vectors for backgrounds b 0 , b 1 on either side of the guidewire for said sequence of windows, averaged over background widths;

calculating the mean vectors μ N , μ b 0 , μ b 1 as a moving average of the moving window, and calculating the variances σ N , σ B from the foreground and background means; and

calculating the distance from

d

2

⁡

(

w

)

=

1

8

⁢

n

⁢

∑

i

=

1

n

⁢

⁢

[

(

μ

N

i

-

μ

b

⁢

⁢

0

i

)

2

+

(

μ

N

i

-

μ

b

⁢

⁢

1

i

)

2

]

⁢

(

σ

N

+

σ

B

+

ɛ

)

-

1

+

1

2

⁢

log

⁢

1

2

⁢

(

σ

N

+

σ

B

)

σ

N

⁢

σ

B

wherein the sum is over the sequence of windows, and n is the number of windows.

13. A program storage device readable by a computer, tangibly embodying a program of instructions executable by the computer to perform the method steps for online optimization of guidewire visibility in fluoroscopic images, said method comprising the steps of:

providing an digitized image acquired from a fluoroscopic imaging system, said image comprising an array of intensities corresponding to a 2-dimensional grid of pixels;

detecting a guidewire in said fluoroscopic image;

enhancing the visibility of the guidewire in said fluoroscopic image;

calculating a visibility measure of said guidewire in said fluoroscopic image; and

readjusting acquisition parameters of said fluoroscopic imaging system wherein said guidewire visibility is improved,

wherein enhancing guidewire visibility comprises constructing a mask image I M wherein pixels that have high filter response and are at most one pixel from said guidewire are also given value 1 and value 0 elsewhere, mask I M over every 3×3 pixel neighborhood, and multiplying the blurred mask by 0.1×(max(I)−min(I)) and subtracting the result from said fluoroscopic image.

14. The computer readable program storage device of claim 13 , wherein detecting a guidewire comprises filtering said image with one or more steerable filters of different angular orientations, assigning a weight W and orientation O for each pixel based on said filter response for each pixel, wherein each pixel weight is assigned to a function of a maximum filter response magnitude and said pixel orientation is calculated from the angle producing said maximum filter response if said magnitude is greater than zero, wherein guidewire pixels have a higher weight than non-guidewire pixels.

15. The computer readable program storage device of claim 13 , wherein blurring said mask over a 3×3 pixel neighborhood comprises multiplying 3×3 sub-matrices of said mask by the following 3×3 matrix:

[

0.0625

0.1250

0.0625

0.1250

0.2500

0.1250

0.0625

0.1250

0.0625

]

16. The computer readable program storage device of claim 13 , wherein calculating a visibility measure comprises:

selecting a set of pixels on said guidewire;

selecting a set of background pixels in a neighborhood of said guidewire that do not include the guidewire pixels;

calculating a mean intensity μ N , μ B , and intensity variance σ N , σ B for both the guidewire pixels and the background pixels, respectively;

calculating a distance measure d 2 between said 2 sets of pixels as

d

2

⁡

(

w

)

=

1

4

⁢

(

σ

N

-

σ

B

)

2

⁢

(

σ

N

+

σ

B

)

-

1

+

1

2

⁢

log

⁢

1

2

⁢

(

σ

N

+

σ

B

)

σ

N

⁢

σ

B

,

wherein w represents the guidewire pixels; and

calculating a visibility V of said guidewire from

V

=

1

n

⁢

∑

i

=

1

n

⁢

⁢

d

2

⁡

(

w

i

)

,

wherein the sum is over the guidewire pixels.

17. The computer readable program storage device of claim 16 , the method further comprising adding a contrast term ε to said distance measure:

d

2

⁡

(

w

)

=

1

4

⁢

(

μ

N

-

μ

B

)

2

⁢

(

σ

N

+

σ

B

+

ɛ

)

-

1

+

1

2

⁢

log

⁢

1

2

⁢

(

σ

N

+

σ

B

)

σ

N

⁢

σ

B

.

wherein a relative importance of the two terms is tuned by a value of said ε-term.

18. The computer readable program storage device of claim 17 , wherein calculating a distance measure between said 2 sets of pixels comprises:

normalizing the fluroscopic image to take on values in the range from 0 to 1;

blurring said image with a Gaussian to remove high frequency noise;

computing a pixel value vector for the guidewire for a sequence of windows centered about and normal to the guidewire, averaged over a wire width;

computing pixel value vectors for backgrounds b 0 , b 1 on either side of the guidewire for said sequence of windows, averaged over background widths;

calculating the mean vectors μ N , μ b 0 , μ b 1 as a moving average of the moving window, and calculating the variances σ N , σ B from the foreground and background means; and

calculating the distance from

d

2

⁡

(

w

)

=

1

8

⁢

⁢

n

⁢

∑

i

=

1

n

⁢

[

(

μ

N

i

-

μ

b

⁢

⁢

0

i

)

2

+

(

μ

N

i

-

μ

b

⁢

⁢

1

i

)

2

]

⁢

(

σ

N

+

σ

B

+

ɛ

)

-

1

+

1

2

⁢

log

⁢

1

2

⁢

(

σ

N

+

σ

B

)

σ

N

⁢

σ

B

wherein the sum is over the sequence of windows, and n is the number of windows.