MASS SEGMENTATION USING MIRROR IMAGE OF REGION OF INTEREST

• US 20100226551A1
• Filed: 05/18/2010
• Published: 09/09/2010
• Est. Priority Date: 12/21/2006
• Status: Active Grant

• Current Assignee: Chao Shi, Daniel Russakoff
• Original Assignee: Chao Shi, Daniel Russakoff
• Inventors: SHI, Chao, Russakoff, Daniel
• Granted Patent: US 7,936,924 B2
• Time in Patent Office: Days
• US Class Current: 382/128
• CPC Class Codes: G06V 10/25 Determination of region of interest [ROI] or a volume of interest [VOI]
• CPC Class Codes: G06V 10/755 Deformable models or variational models, e.g. snakes or active contours
• CPC Class Codes: G06V 2201/03 Recognition of patterns in medical or anatomical images

Abstract

A method and an apparatus for automatic segmentation of an image representing a mass of a tissue region based on dynamic programming that guarantees an accurate and closed contour of the mass is disclosed. The method according to one embodiment accesses digital image data representing an image including the mass of the tissue region, creates a mirror image of the digital image data, extracts a Region of Interest (ROI) which includes a portion of the mirror image containing the mass, transforms the ROI to polar space for obtaining a polar image of the ROI, assigns local cost to sub portions of the polar image, and finds a contour of the mass based on the assigned local cost.

Claims

1. A method for segmenting an image representing a mass of a tissue, comprising:

accessing digital image data representing an image including a tissue region;

extracting from said digital image data a Region of Interest (ROI) surrounding the mass;

transforming the ROI to polar space for obtaining a polar image of the ROI;

calculating an edge strength of the mass from said polar image;

calculating an expected gray level which corresponds with an edge of the mass by using Gaussian mixture;

calculating an expected mass radius by using data from said expected gray level;

assigning local cost to sub portions of said polar image as a weighted combination of the edge strength, expected gray level, and expected mass radius, wherein weight of the edge strength is larger in value than weight of the expected gray level; and

finding a contour of the mass based on the assigned local costs.

2. The method according to claim 1, further comprising:

assigning local cost to each pixel in the polar image in accordance with the following cost function;

c(i,j)=w_ss(i,j)+w_gg(i,j)+w_rr(i,j)

where s(i, j) represents the edge strength, g(i, j) is the deviation from an expected gray level and r(i, j) is the deviation from an expected mass radius, and w_s, w_g, and w_r represent the weights of the components.

3. The method according to claim 2, further comprising:

defining the edge strength as;

$s<br><br>\left(i,j\right)=\frac{\max <br><br>\left(y^{\mathrm{\prime <br><br>}}\right)-y^{\mathrm{\prime <br><br>}}<br><br>\left(i,j\right)}{\max <br><br>\left(y^{\mathrm{\prime <br><br>}}\right)}$

where y′

is the gradient magnitude in vertical direction in the polar image.

4. The method according to claim 2, further comprising:

defining the deviation from the expected gray level as;

g(i,j)=abs[G(i,j)−

g]

where G(i, j) is the intensity value of the pixel (i, j) and g is the expected gray level.

5. The method according to claim 4, further comprising:

normalizing the deviation from the expected gray level into a range of [−

1, 1].

6. The method according to claim 2, further comprising:

defining the deviation from the expected mass radius as;

$r<br><br>\left(i,j\right)=\{\begin{array}{cc}\begin{array}{c}1.0-\frac{1}{\sqrt{2<br><br>\mathrm{\pi <br><br>}<br><br><br><br>{\mathrm{\sigma <br><br>}}_1^2}}<br><br>\exp \\ \frac{\left(-\frac{{\left(j+{\mathrm{\sigma <br><br>}}_1-R\right)}^2}{2<br><br>{\mathrm{\sigma <br><br>}}_1^2}\right)}{g<br><br><br><br>\max }\end{array}& \mathrm{when}<br><br><br><br>\left(j<<br><br>R-{\mathrm{\sigma <br><br>}}_1\right)\\ 0.0& \mathrm{when}<br><br><br><br>\left(R-{\mathrm{\sigma <br><br>}}_1<br><br>j\le <br><br>R+{\mathrm{\sigma <br><br>}}_1\right)\\ \begin{array}{c}1.0-\frac{1}{\sqrt{2<br><br>\mathrm{\pi <br><br>}<br><br><br><br>{\mathrm{\sigma <br><br>}}_1^2<br><br>}}<br><br>\exp \\ \frac{\left(-\frac{{\left(j-{\mathrm{\sigma <br><br>}}_1-R\right)}^2}{2<br><br>{\mathrm{\sigma <br><br>}}_1^2}\right)}{g<br><br><br><br>\max <br><br>}\end{array}& \mathrm{when}<br><br><br><br>\left(j><br><br>R+{\mathrm{\sigma <br><br>}}_1\right)\end{array}$

where σ

represents standard deviation, R represents expected mass radius, σ

₁=1.5σ

, and

$g<br><br><br><br>\max =\frac{1}{\sqrt{2<br><br>\mathrm{\pi <br><br>}<br><br><br><br>{\mathrm{\sigma <br><br>}}_1^2}}.$

7. The method according to claim 1, further comprising:

calculating the expected gray level with histogram, wherein the expected gray level is the gray level which corresponds with the edge of the mass.

8. The method according to claim 1, wherein the weights of the expected gray level and the expected radius are approximately the same in value.

9. The method according to claim 1, wherein the ratio for the weights of the edge strength, the expected gray level and the expected radius is approximately 1:

0.25;

0.25.

10. The method according to claim 1, further comprising:

constructing a cumulative cost matrix as a function of the local cost.

11. The method according to claim 10, further comprising:

extending the cost matrix by plotting the local cost in an interval from −

3π

to 3π

.

12. An apparatus for automatic segmentation of an image representing a mass of a tissue region, comprising:

an image data input unit for accessing digital image data representing an image including said mass of the tissue region;

a Region of Interest (ROI) extraction unit for extracting a ROI surrounding the mass;

an image transforming unit for transforming said ROI to polar space for obtaining a polar image of the ROI;

an edge detection unit for calculating an edge strength of the mass from said polar image;

a gray level calculation unit for calculating an expected gray level which corresponds with the edge of the mass by using Gaussian mixture;

a mass radius calculation unit for calculating an expected mass radius by using data from said expected gray level;

a local cost assignment unit for assigning local cost to sub portions of said polar image as a weighted combination of said edge strength, said expected gray level, and said expected mass radius, wherein the weight of the edge strength is larger in value than the weight of the expected gray scale; and

a dynamic programming unit for finding a contour of the mass based on said assigned local cost.

13. The apparatus according to claim 12, wherein said local cost assignment unit assigns local cost based on the following cost function:

c(i,j)=w_ss(i,j)+w_gg(i,j)+w_rr(i,j)

where s(i, j) represents the edge strength, g(i, j) is the deviation from an expected gray level and r(i, j) is the deviation from an expected mass radius, and w_s, w_g, and w_r represent the weights of local cost components.

14. The apparatus according to claim 13, wherein said edge detection unit defines the edge strength as:

$s<br><br>\left(i,j\right)=\frac{\max <br><br>\left(y^{\mathrm{\prime <br><br>}}\right)-y^{\mathrm{\prime <br><br>}}<br><br>\left(i,j\right)}{\max <br><br>\left(y^{\mathrm{\prime <br><br>}}\right)}$

where y′

is the gradient magnitude in vertical direction in the polar image.

15. The apparatus according to claim 13, wherein said gray level calculation unit calculates the deviation from the expected gray level as:

g(i,j)=abs[G(i,j)−

g]

where G(i, j) is the intensity value of the pixel (i, j) and g is the expected gray level.

16. The apparatus according to claim 13, wherein said mass radius calculation unit defines the deviation from the expected mass radius as:

$r<br><br>\left(i,j\right)=\{\begin{array}{cc}\begin{array}{c}1.0-\frac{1}{\sqrt{2<br><br>{\mathrm{\pi <br><br>\sigma <br><br>}}_1^2}}<br><br>\exp \\ \frac{\left(-\frac{{\left(j+{\mathrm{\sigma <br><br>}}_1-R\right)}^2}{2<br><br>{\mathrm{\sigma <br><br>}}_1^2}\right)}{g<br><br><br><br>\max }\end{array}& \mathrm{when}<br><br><br><br>\left(j<<br><br>R-{\mathrm{\sigma <br><br>}}_1\right)\\ 0.0& \mathrm{when}<br><br><br><br>\left(R-{\mathrm{\sigma <br><br>}}_1\le <br><br>j\le <br><br>R+{\mathrm{\sigma <br><br>}}_1\right)\\ \begin{array}{c}1.0-\frac{1}{\sqrt{2<br><br>\mathrm{\pi <br><br>}<br><br><br><br>{\mathrm{\sigma <br><br>}}_1^2}}<br><br>\exp \\ \frac{\left(-\frac{{\left(j-{\mathrm{\sigma <br><br>}}_1-R\right)}^2}{2<br><br>{\mathrm{\sigma <br><br>}}_1^2}\right)}{g<br><br><br><br>\max }\end{array}& \mathrm{when}<br><br><br><br>\left(j><br><br>R+{\mathrm{\sigma <br><br>}}_1\right)\end{array}$

where σ

represents standard deviation, R represents expected mass radius, σ

₁=1.5σ

, and

$g<br><br><br><br>\max =\frac{1}{\sqrt{2<br><br>\mathrm{\pi <br><br>}<br><br><br><br>{\mathrm{\sigma <br><br>}}_1^2}}.$

17. A computer readable storage medium having stored thereon computer executable program for segmenting an image representing a mass of a tissue, the computer program when executed causes a processor to execute steps of:

accessing digital image data representing an image including a tissue region;

extracting from said digital image data a Region of Interest (ROI) surrounding the mass;

transforming the ROI to polar space for obtaining a polar image of the ROI;

calculating an edge strength of the mass from said polar image;

calculating an expected gray level which corresponds with an edge of the mass by using Gaussian mixture;

calculating an expected mass radius by using data from said expected gray level;

assigning local cost to sub portions of said polar image as a weighted combination of the edge strength, expected gray level, and expected mass radius, wherein weight of the edge strength is larger in value than weight of the expected gray level; and

finding a contour of the mass based on the assigned local costs.