Method for x-ray antiscatter grid detection and suppression in digital radiography

• US 6,269,176 B1
• Filed: 12/21/1998
• Issued: 07/31/2001
• Est. Priority Date: 12/21/1998
• Status: Expired due to Fees

• Current Assignee: CareStream Health Incorporated
• Original Assignee: Eastman Kodak Company
• Inventors: Wang, Xiaohui, Barski, Lori L.
• Primary Examiner(s): Mancuso, Joseph
• Assistant Examiner(s): Tabatabai, Abolfazl
• Application Number:
  US09/218,136
• Time in Patent Office: 953 Days
• Field of Search: 128/916, 382/106, 382/132, 382/128, 364/413.13, 378/98.3
• US Class Current: 382/128
• 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 2207/20056 Discrete and fast Fourier transform, [DFT, FFT]
• CPC Class Codes: G06T 5/10 using non-spatial domain filtering
• CPC Class Codes: G06T 5/70 Denoising; Smoothing
• CPC Class Codes: Y10S 128/922 including image analysis

Abstract

We demonstrate a comprehensive image processing system to automatically detect the use of a stationary anti-scatter device (also referred to as a linear grid) and suppress the artifacts that such a device creates in a digital radiographic images. The detection process consists of the stages of determining an appropriate region of the image from which to perform the analysis; performing a spectral analysis to identify the grid frequencies; and identifying the most likely grid line frequency. When a grid is detected, the results of the detection process are used to perform adaptive grid suppression using custom designed blurring filters and a plurality of adaptation methods.

Claims

1. A digital medical imaging method for improving the presentation of digital projection radiography images comprising the steps of:

providing an input digital radiographic image;

determining a plurality of power spectra from said input digital radiographic image;

identifying all candidate grid frequencies and attributes from said power spectra;

selecting the most likely grid frequencies;

orienting the image for filtering;

calculating the required kernel size as a function of pixel spacing, grid frequency, grid energy and grid half-width full modulation value;

calculating the gaussian sigma value from which to start generating filtering kernels;

building the bank of gaussian filters;

adaptive filtering the image using the bank of gaussian filters; and

reorienting the image to original orientation.

2. The method of claim 1 wherein said providing step includes converting an x-ray image stored in a storage phosphor into said input digital radiographic image.

3. The method of claim 1 wherein said providing step includes digitizing a radiographic image on x-ray film to produce said input digital radiographic image.

4. The method of claim 1 wherein said providing step includes converting an x-ray image stored in a flat-panel detector into said input digital radiographic image.

5. The method of claim 1 wherein said power spectra are obtained using FET.

6. The method of claim 1 wherein said determining step obtains said power spectra from the whole input image.

7. The method of claim 1 wherein said determining step obtains said power spectra from a partial region of the input image.

8. The method of claim 7 wherein said partial region includes the most active region of grid lines.

9. The method of claim 8 wherein said most active region of grid lines is obtained based on the local pixel mean.

10. The method of claim 8 wherein said most active region of grid lines is determined using steps which include:

selecting the size of a square window for analysis;

for each location of the square window within the original image, calculating the mean of all the pixels within the square; and

choosing the square window location corresponding to the minimum mean for a “

white-bone”

original image, or the location corresponding to the maximum mean for a “

black-bone”

original image as the most active region of grid lines.

11. The method of claim 8 wherein said most active region of grid lines is determined using steps which include:

sub-sampling the original image;

selecting the size of a square window for the analysis of the sub-sampled image;

for each location of the square window within the sub-sampled image, calculating the mean of all the pixels within the square;

choosing the square window location corresponding to the minimum mean for a “

white-bone”

sub-sampled image, or the location corresponding to the maximum mean for a “

black-bone”

sub-sampled image; and

mapping the chosen square window location and the window itself to the original image as the most active region of grid lines.

12. A method of claim 1 wherein said determining step includes:

calculating the 1D power spectra of a selected number of rows and columns of pixels of the original image;

averaging the 1D power spectra of all the selected rows and columns, respectively; and

taking the resulting two averaged 1D power spectra as said spectra in claim 1.

13. A method of claim 7 wherein said determining step includes:

calculating the 1D power spectra of a selected number of rows and columns of pixels within said partial region;

averaging the 1D power spectra of all the selected rows and columns, respectively; and

taking the resulting two averaged 1D power spectra as said spectra in claim 7.

14. A method of claim 1 wherein said determining step includes:

calculating the 2D power spectrum of the input image;

averaging a selected number of rows and columns of the 2D power spectrum, respectively; and

taking the resulting two averaged 1D profile as said spectra in claim 1.

15. A method of claim 7 wherein said determining step includes:

calculating the 2D power spectrum of said partial region;

averaging a selected number of rows and columns of the 2D power spectrum, respectively; and

taking the resulting two averaged 1D profile as said spectra in claim 7.

16. The method of claim 1 wherein said identifying step includes the steps of:

optionally smoothing said power spectra;

optionally using morphological opening filter for power spectrum background subtraction;

identifying all the local maxima in said power spectra; and

obtaining the attributes of each local maximum, which include frequency, half-width of full maximum, total energy, and peak magnitude.

17. The method of claim 16 wherein said power spectra are two 1D power spectra, one in x direction and the other in y direction.

18. The method of claim 16 wherein said power spectra is actually a 2D power spectrum.

19. The method of claim 1 wherein said selecting step includes the steps of:

eliminating those peaks whose peak magnitude is smaller than a predetermined threshold;

eliminating those peaks whose frequency is lower than a predetermined threshold;

eliminating those peaks whose energy is less than a predetermined threshold,

calculating the Figure of Merits (FOMs) of all the rest peaks, which include frequency FOM, energy coherence FOM and total FOM;

eliminating those peaks whose fom_tot is lower than a predetermined threshold; and

ranking the rest of the peaks based on their fom_tot.

20. The method of claim 1 wherein said orienting step includes the steps of transposing the image such that grid lines are perpendicular to horizontally oriented 1D kernels.

21. The method of claim 1 wherein said orienting step includes the steps of setting orientation flag such that 1D kernels are oriented perpendicular to grid lines in input image.

22. The method of claim 1 wherein said calculating the required kernel size step includes the steps of calculating the width of a grid shadow, calculating the kernel size, setting the kernel size to the to the half-width full modulation value if said kernel size is smaller, making the kernel size odd, initializing sigma and sigma scale factor;

calculating the grid energy scale factor, and setting the first kernel sigma to the product of the initial sigma and the grid energy scale factor.

23. The method of claim 1 wherein said building step includes the iterative steps of using the 1D equation of the gaussian distribution to calculate kernel values, and multiplying sigma by the sigma scale factor until the required number of kernels have been built.

24. The method of claim 1 wherein said adaptive filtering step includes the steps of finding the useful dynamic range, linearly quantizing the dynamic range to build a kernel look-up table (LUT), selecting the kernel from the bank of kernels based on the kernel LUT which is indexed by the code value of the pixel being processed, and convolving said kernel over the region corresponding to the pixel being processed.

25. The method of claim 1 wherein said adaptive filtering step includes the steps of finding the useful dynamic range, quantizing the dynamic range based upon perceptual linearization to build a kernel look-up table (LUT), selecting the kernel from the bank of kernels based on the kernel LUT which is indexed by the code value of the pixel being processed, and convolving said kernel over the region corresponding to the pixel being processed.

26. The method of claim 1 wherein said adaptive filtering step includes the steps of finding the useful dynamic range, quantizing the dynamic range, calculating image activity histogram, using both the code value of the pixel being processed as well as the amount of image activity of said pixel to select kernel LUT, and convolving said kernel over the region corresponding to the pixel being processed.

27. The method of claim 26 wherein said calculating step includes selecting an as the expected mean of the range, accumulating grid noise estimates using a sliding window where at each location of the window, the range is measured and the standard deviation estimated, storing said standard deviations as a function of code value and averaging stored standard deviations.

28. The method of claim 1 wherein said adaptive filter step includes the iterative steps of selecting a kernel of low blurring capability, convolving said filter on entire image, reprocessing the resultant image with the grid detection process including said determining step, said identifying step, and said selecting step, testing if the resulting grid energy is below the required threshold, and repeating said steps until grid energy falls below said threshold.