METHOD OF SUPPRESSING OBSCURING FEATURES IN AN IMAGE

• US 20090214099A1
• Filed: 02/24/2009
• Published: 08/27/2009
• Est. Priority Date: 02/27/2008
• Status: Active Grant

• Current Assignee: Siemens Computer Aided Diagnosis Ltd. (Siemens AG)
• Original Assignee: Siemens Computer Aided Diagnosis Ltd. (Siemens AG)
• Inventors: Merlet, Nicolas J.
• Granted Patent: US 8,233,692 B2
• Time in Patent Office: Days
• US Class Current: 382/132
• CPC Class Codes: G06T 2207/10116 X-ray image
• CPC Class Codes: G06T 2207/30061 Lung
• CPC Class Codes: G06T 5/30 Erosion or dilatation, e.g. thinning
• CPC Class Codes: G06T 5/77 Retouching; Inpainting; Scratch removal
• CPC Class Codes: G06T 7/11 Region-based segmentation

Abstract

A method of removing obscuring features of relatively standard size and shape from an image comprising the obscuring features, features of interest and background, comprising the steps of applying a geometric transformation to generate a transformed image with the obscuring features aligned with an axis of image; applying an attenuation algorithm to the transformed image to suppress features having a dimension in direction of axis of range appropriate to dimension of the obscuring features in similar alignment, and applying a reverse geometrical transformation to produce a modified image with obscuring features attenuated.

Claims

1. A method of removing obscuring features of relatively standard size and shape from an image comprising the obscuring features, features of interest and background, comprising:

b) selecting Cartesian axes;

c) applying a geometric transformation to generate a transformed image with the obscuring features substantially aligned with one of said axes of image;

d) applying an attenuation algorithm to the transformed image to suppress features having a dimension in direction of one of said axes of range appropriate to dimension of the distracting features in same direction; and

e) applying a reverse geometrical transformation to produce a modified image with distracting features attenuated.

2. The method of claim 1 comprising a preliminary stage (a) of segmenting a region of interest.

3. The method of claim 1 wherein said image is a medical image of a thoracic cavity and said obscuring features are rib depictions and said features or interest and background are lung tissue and malignancies, wherein step (c) comprises the steps of applying a geometric transformation to align the rib depictions with a horizontal axis of the medical image;

step (d) comprises applying an attenuation algorithm to suppress the rib depictions; and

step (e) comprises applying a reverse geometrical transformation, thereby generating a modified image wherein the rib depictions are suppressed.

4. The method of claim 3 wherein the image is an x-ray radiograph.

5. The method of claim 3 comprising a preliminary stage (a) of segmenting contour of lung.

6. The method of claim 3 wherein step (c) of applying a geometrical transformation comprises a step (c1) of applying a coarse transformation.

7. The method of claim 6 wherein step (c) further comprises a subsequent step (c2) of applying a fine transformation.

8. The method of claim 7 wherein step (c1) comprises the following substeps:

c(i) selecting a section of a first column of a base Image I₀ between two contour points;

c(ii) fitting a polynomial approximation to distance vs. grey level along the column;

c(iii) generating a difference profile by plotting the difference between the value of the grayscale profile and the value of the polynomial approximation to the grayscale profile for each pixel along the section of the column;

c(iv) smoothing the difference profile using a smoothing function;

c(v) repeating steps c(i) to c(iv) to produce a smoothed-difference-profile for a second column of the base Image I₀;

c(vi) correlating the smoothed-difference-profiles of the first and second columns by applying an affine transformation to the smoothed-difference-profile of the second column;

c(vii) repeating steps c(i) to c(iv) to produce a smoothed-difference-profile for a next column of the base Image I₀;

c(viii) correlating the smoothed difference profiles of the next and previous columns by applying an affine transformation to the smoothed-difference-profile of the next column so as to minimize the correlation-difference between the previous transformed smoothed-difference-profile and the next transformed smoothed-difference-profile; and

c(ix) repeating until an extremity of the lung is reached.

9. The method of claim 8 wherein the section of column in sub-step c(i) is distance between contour points minus a safety margin.

10. The method of claim 9 wherein the safety margin is vertically displaced by 5% from each contour point.

11. The method of claim 8 comprising a limitation selected from the list including:

(I) the polynomial approximation of step c(iii) being a 5^th order polynomial;

(II) the smoothing function of step c(iv) being a sigmoid transformation;

(III) the affine transformation of step c(vi) comprising applying a linear transformation of type x′

→

ax+b where a defines a degree of stretch and b defines an offset, thereby minimizing correlation-difference between the first smoothed-difference-profile and the transformed second smoothed-difference-profile;

(IV) the first column is a median x coordinate of a lung and correlation is performed separately towards left and right sides of the lung;

(V) storing cumulative affine transform coefficients for each column for use in inverse geometric transform for converting the attenuated transformed image I₅ into a desired final image I_f; and

(VI) reaching an extremity of a lung is defined as reaching point where column length does not exceed 100 pixels.

12. The method of claim 3 wherein stage (d) of suppressing the ribs comprises the following substeps:

d(i) identifying short horizontal elements having horizontal dimensions too short to be ribs, and

d(ii) discounting said short horizontal elements to identify suspect ribs by elimination.

13. The method of claim 11 further comprising the step d(iii) of identifying suspect ribs as ribs by examination of vertical dimensions in transformed image.

14. The method of claim 11 further comprising the substeps of removing suspected ribs from the second transformed image I₂ by plotting the grayscale of each column to form an irregular convex shaped profile, and

enveloping the irregular convex shaped profile using a convexity transform.

15. The method of claim 11 wherein the step d(i) of identifying short horizontal elements comprises:

plotting grayscale against displacement for each row to generate an irregular convex shaped profile;

applying a convexity fit to envelope the irregular convex shaped profile with a hull; and

producing a second transformed image I₂ by replacing the grayscale profiles of each row of I₁ with the associated approximate hull-envelope.

16. The method of claim 15 wherein the step of applying the convexity fit comprises the following steps:

(A) letting the coordinates of each point profile be (x, g) where x corresponds to horizontal-pixel number and g is grayscale value thereof.

(B) taking a first point of the Hull at the point (x₁, g₁);

(C) selecting a second point (x₂, g₂) within the next range of pixels where

[(g_ig-g₁)/(x_i-x₁)] is lowest; and

(D) repeating steps alpha and beta for all subsequent points.

17. The method of claim 16 wherein the range is 200 pixels.

18. A computer program for computer aided diagnosis of lung disorders comprising a sub-procedure including the method of claim 3.

19. A computer comprising:

a processor, data input means, at least one data storage memory, random access memory, a user interface comprising user selectable keys and a visual display unit, said computer configured to receive a medical image having Cartesian axes of a body section with bone and/or cartilage depictions that obscures organs and/or soft tissue, wherein the image processing comprises the following steps;

(a) applying a geometric transformation to align the bone depictions with an axis of the medical image;

(b) applying an attenuation algorithm to suppress the bone and/or cartilage depictions; and

step (c) applying a reverse geometrical transformation, thereby generating a modified image of the body section wherein the bone and/or cartilage depictions are suppressed.

20. The computer of claim 19 wherein the medical image is selected from the list including:

x-ray images of limbs, x-ray images of organs, anterior posterior x-ray images, posterior anterior x-ray images, left lateral x-ray images, right lateral x-ray images, decubitus view x-ray images, lordotic view x-rays, images, expiratory view x-ray images, oblique view x-ray images, NMR images and ultrasound images.