Method and system for localization of inter-rib spaces and automated lung texture analysis in digital chest radiographs

• US 4,851,984 A
• Filed: 08/03/1987
• Issued: 07/25/1989
• Est. Priority Date: 08/03/1987
• Status: Expired due to Term

• Current Assignee: University of Chicago The (University of Chicago)
• Original Assignee: University of Chicago
• Inventors: Doi, Kunio, Katsuragawa, Shigehiko
• Primary Examiner(s): Pellinen, A. D.
• Assistant Examiner(s): Gaffin, Jeffrey A.
• Application Number:
  US07/081,143
• Time in Patent Office: 722 Days
• Field of Search: 364/414, 378/99, 378/90, 378/901, 382/54, 382/52, 382/6, 358/111
• US Class Current: 382/108
• CPC Class Codes: A61B 5/08 Detecting, measuring or recording devices for evaluating the respiratory organs A61B5/0205 takes precedence
• CPC Class Codes: A61B 5/7257 using Fourier transforms
• CPC Class Codes: A61B 6/5258 involving detection or reduction of artifacts or noise
• CPC Class Codes: G06T 2207/30061 Lung
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/194 involving foreground-background segmentation
• CPC Class Codes: Y10S 378/901 Computer tomography program or processor

Abstract

A method and system for automated analysis of digital radiographic images in which regions-of-interest (ROI'"'"'s) are first determined, and subsequently analyzed for abnormalities. To locate the ROI'"'"'s, the outer ribcage and midline boundary locations of the chest image are determined from the digital image data. Vertical profiles are then selected and background trend is then removed from each vertical profile. A shift-variant sinusoidal function is fitted to each vertical profile and ROI'"'"'s are selected based on the fitted functions. The non-uniform background trend is removed from the original image data of each selected ROI to obtain corrected data. The power spectrum of the lung texture is obtained from the 2D Fourier transform of the corrected data and is filtered by the human visual response. Finally, the root-mean-square (rms) variation, R, and the first moment of the power spectrum, M, are determined as quantitative texture measures for the magnitude and coarseness (or fineness), respectively, of the lung texture.

Claims

1. A method for automated analysis of lung texture in a radiographic chest image, comprising:

obtaining digital image data representative of said image;

selecting at least one region of interest (ROI) in said image for analysis;

removing background trend due to the gross anatomy of the lung and chest wall from the image data corresponding to said ROI and thereby producing corrected image data, representative of fluctuating patterns of the underlying lung texture, in which the background trend is removed;

processing the corrected image data to extract predetermined characteristics therefrom for evaluation.

2. The method according to claim 1, wherein said processing step comprises:

deriving objective texture measures based on the power spectrum of the corrected data.

3. The method according to claim 1, wherein said processing step comprises:

producing a two-dimensional (2D) Fourier transform of the corrected image data to obtain Fourier transformed data T(u,v), where u and v are spatial frequencies in a cartesian coordinate system;

bandpass filtering the Fourier transformed data T(u,v) to obtain filtered data;

determining the power spectrum of the filtered data; and

extracting at least one predetermined characteristic from the determined power spectrum.

4. The method according to claim 3, wherein said step of bandpass filtering comprises:

filtering T(u,v) by a human visual response, V(u,v), where ##EQU5## to obtain filtered data (T(u,v)V(u,v)), where u_o and D are predetermined constants.

5. The method according to claim 4, wherein said extracting step comprises:

determining the root-mean-square (RMS) variation, R, of the power spectrum of the filtered data, where ##EQU6##

6. The method according to claim 5, wherein said extracting step comprises:

determining the first moment of the power spectrum, M, of the filtered data, where ##EQU7##

7. The method according to claim 6, wherein said background trend removing step comprises:

fitting a two-dimensional surface to the image data of said at least one ROI and obtaining two-dimensional fitting data corresponding to said surface; and

subtracting the two-dimensional fitting data from said ROI image data.

8. The method according to claim 7, wherein said ROI selecting step comprises:

boundary processing said digital image data to derive ribcage edge and midline boundaries;

deriving at least one vertical profile at predetermined locations in relation to the ribcage edge and midline boundaries derived in said boundary processing step;

removing background trend from the at least one vertical profile to obtain vertical profile corrected (VPC) data; and

selecting said at least one ROI based on said VPC data.

9. The method according to claim 8, wherein said boundary processing step comprises:

obtaining a horizontal signature at a horizontal line at a predetermined vertical location of said chest image from said image data;

determining a maximum value of said horizontal signature and defining the pixel location of that pixel having said maximum value as a midline boundary; and

determining a second derivative of said horizontal signature, detecting at which pixel said second derivative is a minimum, and defining the pixel location at which said second derivative is a minimum as an edge boundary.

10. The method according to claim 9, comprising:

obtaining edge boundaries from plural horizontal lines.

11. The method according to claim 10, wherein said vertical profile determining step comprises:

defining said at least one vertical profile based on pixel values of pixels located a predetermined distance from the edge boundaries determined during the repeated performance of said boundary processing step.

12. The method according to claim 11, wherein said step of selecting said at least one ROI comprises:

fitting a shift-variant sinusoidal function to said VPC data; and

selecting said at least one ROI based on the fitted shift-variant sinusoidal function.

13. The method according to claim 12, comprising:

displaying an image of said image data with the selected ROI superimposed thereon.

14. The method according to claim 13, comprising:

editing the at least one selected ROI by means of user interactive control of said displaying step.

15. The method according to claim 4, wherein said extracting step comprises:

determining the first moment of the power spectrum, M, of the filtered data, where ##EQU8##

16. The method according to claim 1, wherein said background trend removing step comprises:

fitting a two-dimensional surface to the image data of said at least one ROI and obtaining two-dimensional fitting data corresponding to said surface; and

subtracting the two-dimensional fitting data from said ROI image data.

17. The method according to claim 16, wherein said fitting step comprises:

fitting a second-order polynomial surface to said image data.

18. The method according to claim 1, wherein said ROI selecting step comprises:

boundary processing said digital image data to derive ribcage edge and midline boundaries;

deriving at least one vertical profile at a predetermined locations in relation to the ribcage edge and midline boundaries derived in said boundary processing step;

removing background trend from the at least one vertical profile to obtain vertical profile corrected (VPC) data; and

selecting said at least one ROI based on said VPC data.

19. The method according to claim 18, wherein said boundary processing step comprises:

obtaining a horizontal signature at a horizontal line at a predetermined vertical location of said chest image from said image data;

determining a maximum value of said horizontal signature and defining the pixel location of that pixel having said maximum value as a midline boundary; and

determining a second derivative of said horizontal signature, detecting at which pixel said second derivative is a minimum, and defining the pixel location at which said second derivative is a minimum as an edge boundary.

20. The method according to claim 19, comprising:

obtaining edge boundaries from plural horizontal lines.

21. The method according to claim 20, wherein said vertical profile deriving step comprises:

defining said at least one vertical profile based on pixel values of pixels located a predetermined distance from the edge boundaries determined during the repeated performance of said boundary processing step.

22. The method according to claim 21, wherein said step of selecting said at least one ROI comprises:

fitting a shift-variant sinusoidal function to said VPC data; and

selecting said at least one ROI based on the fitted shift-variant sinusoidal function.

23. The method according to claim 22, comprising:

editing the at least one selected ROI by means of user interactive control of said displaying step.

24. The method according to claim 18, wherein said step of selecting said at least one ROI comprises:

fitting a shift-variant sinusoidal function to said VPC data; and

selecting said at least one ROI based on the fitted shift-variant sinusoidal function.

25. The method according to claim 24, comprising:

displaying an image of said image data with the at least one selected ROI superimposed thereon.

26. A method for selecting at least one region of interest (ROI) for analysis of lung texture in a radiographic chest image, comprising:

obtaining digital image data representative of said image;

boundary processing said digital image data to derive ribcage edge and midline boundaries;

deriving at least one vertical profile at a predetermined location in relation to the ribcage edge and midline boundaries derived in said boundary processing step;

removing background trend from the at least one vertical profile to obtain vertical profile corrected (VPC) data;

selecting said at least one ROI based on said VPC data.

27. The method according to claim 26, wherein said boundary processing step comprises:

obtaining a horizontal signature at a horizontal line at a predetermined vertical location of said chest image from said image data;

determining a maximum value of said horizontal signature and defining the pixel location of that pixel having said maximum value as a midline boundary; and

determining a second derivative of said horizontal signature, detecting at which pixel said second derivative is a minimum, and defining the pixel location at which said second derivative is a minimum as an edge boundary.

28. The method according to claim 27, comprising:

obtaining edge boundaries from plural horizontal lines.

29. The method according to claim 28, wherein said vertical profile determining step comprises

defining said at least one vertical profile based on pixel values of pixel located at predetermined distance from the edge boundaries determined during the repeated performance of said boundary processing step.

30. The method according to claim 29, wherein said step of selecting said at least one ROI comprises:

fitting a shift-variant sinusoidal function to said VPC data; and

selecting said at least one ROI based on the fitted shift-variant sinusoidal function.

31. The method according to claim 26, wherein said step of selecting said at least one ROI comprises:

fitting a shift-variant sinusoidal function to said VPC data; and

selecting said at least one ROI based on the fitted shift-variant sinusoidal function.

32. The method according to claim 31, comprising:

displaying an image of said image data with the selected ROI superimposed thereon.

33. The method according to claim 32, comprising:

editing the at least one selected ROI by means of user interactive control of said displaying step.

34. A system for automated analysis of lung texture in a radiographic chest image, comprising:

means for obtaining digital image data representative of said image;

means for selecting at least one region of interest (ROI) in said image for analysis;

means for removing background trend due to the gross anatomy of the lung and chest wall from the image data corresponding to said ROI and thereby producing corrected image data, representative of fluctuating patterns of the underlying lung texture, in which the background trend is removed;

processing means for extracting from the corrected image data predetermined characteristics for evaluation.

35. The system according to claim 34, wherein said processing means comprises:

means for producing a two-dimensional (2D) Fourier transform of the corrected image data to obtain Fourier transformed data T(u,v);

means for bandpass filtering T(u,v);

means for determining the power spectrum of the filtered data;

means for determining the root-mean-square (RMS) variation, R, of the power spectrum of the Fourier transformed data, where ##EQU9## means for determining the first moment of the power spectrum, M, of the Fourier transformed data, where ##EQU10## where V(u,v) defines the filter characteristic of said bandpass filtering means, and u and v are spatial frequencies in cartesian coordinates.

36. The system according to claim 35, wherein the characteristic V(u,v) of said bandpass filtering means is defined in terms of the human visual response as follows:

##EQU11## where u_o and D are predetermined constants.

37. The system according to claim 36, wherein said background trend removing means comprises:

means for fitting a two-dimensional surface to the image data of said at least one ROI and obtaining two-dimensional fitting data corresponding to said surface; and

means for subtracting the two-dimensional fitting data from said ROI image data.

38. The system according to claim 37, wherein said ROI selecting step comprises:

boundary processing means for determining from said digital image data ribcage edge and midline boundaries;

means for deriving at least one vertical profile at predetermined locations in relation to the ribcage edge and midline boundaries determined by said processing means;

removing background trend from the at least one vertical profile to obtain vertical profile corrected (VPC) data; and

means for selecting said at least one ROI based on said VPC data.

39. The system according to claim 38, wherein said boundary processing means comprises

means for a horizontal signature at a horizontal line at a predetermined vertical location of said chest image from said image data;

means for determining a maximum value of said horizontal signature, wherein the pixel location of that pixel having said maximum value is defined as a midline boundary; and

means for determining a second derivative of said horizontal signature and detecting at which pixel said second derivative is a minimum, wherein the pixel location at which said second derivative is a minimum is defined as an edge boundary.

40. The system according to claim 39, comprising:

means for obtaining edge and midline boundaries for plural horizontal lines.

41. The system according to claim 40, wherein said vertical profile deriving means comprises:

means for determining said at least one vertical profile based on pixel values of pixels located a predetermined distance from the edge boundaries determined for said plural horizontal lines.

42. The system according to claim 41, wherein said means for selecting said at least one ROI comprises:

means for fitting a shift-variant sinusoidal function to said VPC data; and

means for selecting said at least one ROI based on the fitted shift-variant sinusoidal function.

43. The system according to claim 42, comprising:

means for displaying an image of said image data with the selected ROI superimposed thereon.

44. The method according to claim 43, comprising:

user interactive control means for editing the at least one selected ROI.

45. The system according to claim 34, wherein said background trend removing means comprises:

means for fitting a two-dimensional surface to the image data of said at least one ROI and obtaining two-dimensional fitting data corresponding to said surface; and

means for subtracting the two-dimensional fitting data from said ROI image data.

46. The system according to claim 45, wherein said fitting means comprises:

means for fitting a second-order polynomial surface to said image data.

47. The system according to claim 34, wherein said ROI selecting means comprises:

boundary processing means for determining from said digital image data ribcage edge and midline boundaries;

means for deriving at least one vertical profile at a predetermined location in relation to the ribcage edge and midline boundaries determined by said boundary processing means;

means for removing background trend from the at least one vertical profile to obtain vertical profile corrected (VPC) data; and

means for selecting said at least one ROI based on said VPC data.

48. The method according to claim 47, wherein said boundary processing means comprises

means for obtaining a horizontal signature at a horizontal line at a predetermined vertical location of said chest image from said image data;

means for determining a maximum value of said horizontal signature wherein the pixel location of that pixel having said maximum value is defined as a midline boundary; and

means for determining a second derivative of said horizontal signature and detecting at which pixel said second derivative is a minimum, wherein the pixel location at which said second derivative is a minimum is defined as an edge boundary.

49. The system according to claim 48, wherein said boundary processing means comprises:

means for obtaining edge and midline boundaries for plural horizontal lines.

50. The system according to claim 49, wherein said vertical profile deriving means comprises:

means for determining said at least one vertical profile based on pixel values of pixels located a predetermined distance from the edge boundaries determined for said plural horizontal signatures.

51. The system according to claim 50, wherein said means for selecting said at least one ROI comprises:

means for fitting a shift-variant sinusoidal function to said VPC data; and

means for selecting said at least one ROI based on the fitted shift-variant sinusoidal function.

52. The system according to claim 47, wherein said means for selecting said at least one ROI comprises:

means for fitting a shift-variant sinusoidal function to said VPC data; and

means for selecting said at least one ROI based on the fitted shift-variant sinusoidal function.

53. The system according to claim 52, comprising:

means for displaying an image of said image data with the at least one selected ROI superimposed thereon.

54. The system according to claim 53, comprising:

interactive control means for editing the at least one selected ROI.

55. A system for selecting at least one region of interest (ROI) for analysis of lung texture in a radiographic chest image, comprising:

means for obtaining digital image data representative of said image;

boundary processing means for deriving from said digital image data ribcage edge and midline boundaries;

means for deriving at least one vertical profile at a predetermined location in relation to the ribcage edge and midline boundaries derived by said boundary processing means;

means for removing background trend from the at least one vertical profile to obtain vertical profile corrected (VPC) data; and

means for selecting said at least one ROI based on said VPC data.

56. The system according to claim 55, wherein said boundary processing means comprises:

means for obtaining a horizontal signature at a predetermined vertical location of said chest image from said image data;

means for determining a maximum value of said horizontal signature, wherein the pixel location of that pixel having said maximum value is defined as a midline boundary; and

means for determining a second derivative of said horizontal signature and detecting at which pixel said second derivative is a minimum, wherein the pixel location at which said second derivative is a minimum is defined as an edge boundary.

57. The system according to claim 56, comprising:

means for obtaining edge and midline boundaries for plural horizontal lines.

58. The system according to claim 57, wherein said vertical profile determining means comprises:

means for determining said at least one vertical profile based on pixel values of pixels located a predetermined distance from the edge boundaries determined for said plural horizontal lines.

59. The system according to claim 58, wherein said step of selecting said at least one ROI comprises:

means for fitting a shift-variant sinusoidal function to said VPC data; and

means for selecting said at least one ROI based on the fitted shift-variant sinusoidal function.

60. The system according to claim 55, wherein said step of selecting said at least one ROI comprises:

means for fitting a shift-variant sinusoidal function to said VPC data; and

means for selecting said at least one ROI based on the fitted shift-variant sinusoidal function.

61. The system according to claim 60, comprising:

means for displaying an image of said image data with the selected ROI superimposed thereon.

62. The system according to claim 61, comprising:

interactive control means for editing the at least one selected ROI.