Methods and apparatuses for correcting a mammogram with an implant, and for segmenting an implant

• US 8,160,346 B2
• Filed: 06/28/2007
• Issued: 04/17/2012
• Est. Priority Date: 06/30/2006
• Status: Active Grant

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Lavoue, Sophie, Razvan, Iordache, Muller, Serge, Gatesoupe, Pascal, Durand, Jérôme
• Primary Examiner(s): Repko, Jason M
• Assistant Examiner(s): Thomas, Mia M
• Application Number:
  US11/770,101
• Publication Number:
  US 20080002872A1
• Time in Patent Office: 1,755 Days
• Field of Search: 382128-132, 382/173, 382/174, 382/177, 382/260, 382/264, 382/276, 382/277, 378/4, 378/21, 600/407, 600/427
• US Class Current: 382/132
• CPC Class Codes: A61B 6/502 for diagnosis of breast, i.e. mammography
• CPC Class Codes: A61B 6/5252 removing objects from field of view, e.g. removing patient table from a CT image
• CPC Class Codes: G06T 2207/10116 X-ray image
• CPC Class Codes: G06T 2207/30052 Implant; Prosthesis
• CPC Class Codes: G06T 2207/30068 Mammography; Breast
• CPC Class Codes: G06T 2207/30096 Tumor; Lesion
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/155 involving morphological operators

Abstract

The invention relates to a method of correcting a digital mammogram of a breast containing an implant. According to the invention, brightnesses situated at and around the implant peak (PI) in the low frequency component of the image (X_LP) are attenuated (ATT) as compared with a function (P1) that is applied to the other brightnesses, thereby attenuating the implant.

The invention also provides a method of segmenting an implant in a radiographic image, the method being characterized by the following steps: smoothing the image by a noise-reducing lowpass filter, applying a non-directional spatial derivative filter, pixel thresholding, and selecting an outline to define the segmented portion of the image that corresponds to the implant.

Claims

1. A method of segmenting an implant in a radiographic image of a portion of a patient'"'"'s body that contains the implant, the method comprising:

(a) smoothing the image by a lowpass first filter for reducing noise in order to obtain a smoothed image;

(b) applying a non-directional spatial derivative second filter to the smoothed image in order to obtain a filtered image;

(c) hysteresis thresholding pixels of the filtered image in order to obtain a thresholded image, the hysteresis thresholding comprising;

(i) retaining pixels having a brightness value greater than or equal to a high first threshold,

(ii) discarding pixels having a brightness value less than a low second threshold, and

(iii) retaining pixels having a brightness value lying between the high first threshold and the low second threshold providing they are adjacent to at least one pixel having a brightness value greater than or equal to the high first threshold;

(d) selecting an outline obtained in the thresholded image, thereby defining a segmented portion of the image corresponding to the implant;

(e) modifying the low second threshold used during the hysteresis thresholding if the segmented portion obtained does not satisfy a calculated criterion, and reapplying the hysteresis thresholding and electing an outline in order to obtain a new segmented portion of the image; and

if the calculated criterion is not satisfied after a predetermined number of modifications or after lowering the low second threshold to a minimum, then dilatations are performed followed by erosions of increasing size on the thresholded image until a closed outline is obtained defining the segmented portion,

wherein prior to the smoothing at least one anisotropic filtering is applied to the image in order to decrease noise, and

wherein the anisotropic filtering applies a filter to the preceding pixels of the image, which filter delivers at the following iteration for the pixel of coordinates s, a new pixel of coordinates s, defined by the equation;

u(s,t+Δ

t)=u(s,t)+Δ

t*div (f (∥

∇

u(s,t)∥

)·

∇

u(s,t))

where;

∇

designates the gradient operator;

div designates the divergence operator;

f is a decreasing non-uniformity function; and

Δ

t is a positive and non-zero predetermined increment.

2. The method according to claim 1, wherein the image smoothing is performed by a Gaussian lowpass first filter.

3. The method according to claim 1, wherein the non-directional spatial derivative second filter applied to the smoothed image is a Laplacian filter.

4. The method according to claim 1, wherein the image smoothing is performed by a Gaussian lowpass first filter, and the non-directional spatial derivative second filter applied to the smoothed image is a Laplacian filter, the Gaussian first filter and the Laplacian second filter together forming a global LoG filter implementing the following function:

5. The method according to claim 1, wherein zero-crossings are looked for in the filtered image, as subjected to the gradient norm calculated on said filtered image, in order to preselect at least one outline.

6. The method according to claim 1, wherein the criterion is at least one of the group constituted by:

the area of the segmented portion belongs to a range of areas relative to the estimated area of the portion of the body in the image; and

the variance of the segmented portion is below a variance threshold.

7. The method according to claim 1, wherein the selecting at least one outline obtained in the thresholded image is performed automatically by comparison with region models defined by prerecorded reference functions.

8. The method according to claim 1, wherein processing is applied to the resulting segmented image portion, which processing consists in applying a mask for blocking out said segmented portion from the initial image.

9. The method according to claim 1, wherein processing is applied to the resulting segmented portion of the image, which processing consists in identifying its outline in the initial image.

10. The method according to claim 1, wherein prior to the smoothing, a dynamic linearization is applied to the image, for expanding the dynamic range of pale zones relative to the dynamic range of dark zones in the image.

11. The method according to claim 10, wherein the dynamic linearization applies a cumulative distribution function to the histogram of the image.

12. The method according to claim 1 wherein the decreasing non-uniformity function f is a Gaussian function.

13. The method according to claim 12, wherein the new pixel of the following iteration is given by:

14. The method according to claim 1, wherein the predetermined increment Δ

t is greater than 0 and less than 0.25.

15. The method according to claim 1, wherein a mask region is grown inside the boundaries of the resulting segmented portion, the region being grown from at least one selected starting point, by iterating at least one growth step, each growth step including within the mask region those pixels that are adjacent to the region obtained at the preceding iteration and that have brightness values that are sufficiently close to those of the region obtained at the preceding iteration.

16. The method according to claim 15, wherein the starting point for growth of the mask region is obtained by:

simple thresholding of the image above a brightness threshold that is less than the maximum brightness of the pixels of the image so as to obtain a first image zone; and

eroding and then dilating the first zone so as to obtain a second zone, with the center of gravity of the second zone then being taken as the starting point for growing the mask region.

17. A method of segmenting an implant and for correcting a digital mammogram of a breast containing an implant, information being represented in the mammogram by pixels of brightness greater than or less than a background, the method comprising:

detecting, in the histogram of brightnesses of the mammogram, an adipose first peak, calculating a first brightness value corresponding to the adipose first peak, detecting a second peak in the histogram of brightnesses of the mammogram, said second peak corresponding to the presence of the implant, and calculating a second brightness value corresponding to the implant second peak;

resolving the mammogram into a low frequency image component and a high frequency image component;

applying a first correction function to the low frequency image component, in order to obtain a processed low frequency image component, the first correction function comprising;

in a first configuration where breast information is represented in the mammogram by pixels of brightness greater than that of a background, an implant attenuation for brightnesses situated at least at and around the implant peak relative to a first function applied to the other brightnesses situated between the first value and the second value;

a second configuration in which breast information is represented in the mammogram by pixels having brightnesses lower than that of a background, an implant amplification for brightnesses situated at least at and around the implant peak relative to a first function applied to the other brightnesses situated between the first value and the second value; and

restoring the corrected mammogram on the basis of the high frequency component and of the processed low frequency component.

18. The method according to claim 17, wherein:

in the first configuration, the attenuation of the first correction function is applied to the brightnesses of the low frequency image component that are greater than or equal to a predetermined third brightness value lying between the first brightness value of the adipose peak and the second brightness value of the implant peak; and

in the second configuration, the amplification of the first correction function is applied to the brightnesses of the low frequency image component that are lower than or equal to a predetermined third brightness value lying between the first brightness value of the adipose peak and the second brightness value of the implant peak.

19. The method according to claim 18, wherein the third brightness value is calculated as corresponding to the lowest valley detected in the brightness histogram of the mammogram, for brightness values that are less than the second brightness value of the implant peak in the first configuration and that are greater than the second brightness value of the implant peak in the second configuration.

20. The method according to claim 18, wherein:

in the first configuration, the attenuation of the first correction function for the low frequency component comprises a decreasing portion for brightnesses lying between the third brightness value and the second brightness value of the implant peak, and a substantially constant portion for brightness values greater than or equal to the second brightness value of the implant peak; and

in the second configuration, the amplification of the first correction function of the low frequency component comprises a decreasing portion for brightnesses lying between the third brightness value and the second brightness value of the implant peak, and a substantially constant portion for brightness values less than or equal to the second brightness value of the implant peak.

21. The method according to claim 20, wherein:

in the first configuration, the attenuation of the first correction function for the low frequency component includes a substantially constant portion equal to the mean brightness value of the adipose peak divided by the mean brightness value of the implant peak; and

in the second configuration, the amplification of the first correction function for the low frequency component includes a substantially constant portion qual to the mean brightness value of the adipose peak (F) divided by the mean brightness value of the implant peak.

22. The method according to claim 17, wherein said first correction function for the low frequency component includes a plateau multiplying brightness values lying between the first brightness value of the adipose peak and the third brightness value by a coefficient that is substantially constant.

23. The method according to claim 17, wherein the first correction function for the low frequency image component includes:

in the first configuration, a raising for brightness values that are less than the first value of the adipose peak; and

in the second configuration, additional attenuation for brightness values greater than the first value of the adipose peak.

24. The method according to claim 23, wherein:

in the first configuration, the raising is a decreasing function for brightness values less than the first value of the adipose peak; and

in the second configuration, the additional attenuation is a decreasing function for brightness values greater than the first value of the adipose peak.

25. The method according to claim 17, wherein a second correction function is applied to the high frequency image component in order to obtain a processed high frequency image component, the second correction function comprising:

in the first configuration, an attenuation for brightness value less than or equal to the first brightness value of the adipose peak relative to a second function applied to brightness values greater than said first brightness value; and

in the second configuration, an amplification for brightness values greater than or equal to the first brightness value of the adipose peak relative to a second function applied to brightness values that are less than said first brightness value.

26. An apparatus for segmenting an implant and for correcting a mammogram of a breast containing an implant, the apparatus comprising:

means for detecting an adipose first peak in a histogram of pixel brightness of the mammogram;

means for calculating a first brightness value corresponding to the adipose first peak;

means for detecting an implant second peak in the histogram of pixel brightness of the mammogram;

means for calculating a second brightness value corresponding to the implant second peak;

means for separating the mammogram into a low frequency image component and into a high frequency image component;

means for applying a correction function to the low frequency image component in order to obtain a processed low frequency image component; and

means for combining the high frequency component and the processed low frequency component into a corrected mammogram.

27. The apparatus according to claim 26 wherein the correction function attenuates the brightness of pixels at and around the implant second peak in a mammogram where breast information is represented by pixels having a brightness greater than that of a background of the mammogram.

28. The apparatus according to claim 26 wherein the correction function amplifies the brightness of pixels at and around the implant second peak in a mammogram where breast information is represented by pixels having a brightness lower than that of a background of the mammogram.