Automated method and system for the detection of gross abnormalities and asymmetries in chest images

• US 5,638,458 A
• Filed: 11/30/1993
• Issued: 06/10/1997
• Est. Priority Date: 11/30/1993
• Status: Expired due to Term

• Current Assignee: ARCH Development Corporation (University of Chicago)
• Original Assignee: ARCH Development Corporation (University of Chicago)
• Inventors: Giger, Maryellen L., MacMahon, Heber, Armato, Samuel
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): Tran, Phuoc
• Application Number:
  US08/159,133
• Time in Patent Office: 1,288 Days
• Field of Search: 382/6, 382/18, 382/19, 382/22, 382/28, 382/51, 382/128, 382/130, 382/131, 382/132, 382/172, 382/190, 382/195, 382/199, 382/206, 382/257, 382/266, 382/270, 382/271, 382/273, 382/28 8, 364/413.13, 364/413.19, 364/413.22, 364/413.23
• 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 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/68 of symmetry
• CPC Class Codes: Y02A 90/10 Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Abstract

A method for the automated detection of gross abnormalities and asymmetries in chest images, including generating image data from radiographic images of the thorax. The image data are then analyzed in order to produce the boundaries of the aerated lung regions in the thorax. This analysis comprises location of the mediastinum and lung apices, iterative global thresholding with centroid testing of contours, local thresholding on regions along initial contours of the aerated lung, correction for regions near the costo- and cardiophrenic angles in the chest, analysis of the areas and density distribution within the aerated lung regions in the chest and determination of the likelihood of the presence of a gross abnormality or asymmetry. Final output could be the computer determined contours of the lungs or the likelihood for abnormality.

Claims

1. A method for detecting an abnormality in a radiographic image, comprising:

obtaining a radiographic image of a subject;

detecting a boundary of a feature in said radiographic image;

determining geometric descriptors of said feature; and

determining whether an abnormality exists using said geometric descriptors;

wherein said step of detecting said boundary of said feature comprises;

performing iterative global gray-level thresholding on said radiographic image, comprising,

determining a plurality of gray-level theshold values

iteratively creating respective binary images defining at least one region based on each of said plurality of said gray-level threshold values,

defining a characteristic limit for said at least one region,

eliminating said at least one region from the respective image and subsequent images iteratively obtained if said at least one region has a characteristic outside said characteristic limit, and

determining a final region resulting after iterative eliminating of regions and iterative thresholding at said plurality of gray-level threshold values as said feature.

2. A method as recited in claim 1, wherein said step of performing iterative global gray-level thresholding comprises:

constructing a gray-level histogram of said radiographic image;

performing a running slope of said gray-level histogram to obtain a smoothed gray-level histogram; and

determining a number of first gray-level threshold values at select locations in said smoothed gray-level histogram.

3. A method as recited in claim 1, wherein said step of iteratively creating said repective binary images comprisies turning on pixels in each iterative image having a gray level less than a respective one of said gray-level threshold values.

4. A method as recited in claim 1, comprising:

said step of defining a characteristic limit comprising defining a centroid limit in said radiographic image; and

said step of eliminating comprising eliminating said region if it has a centroid outside said centroid limit.

5. A method as recited in claim 1, comprising:

performing a morphological open operation on selected ones of said binary images.

6. A method as recited in claim 1, comprising:

using said binary images to create an initial boundary of said feature; and

performing local gray-level thresholding using said initial boundary of said feature.

14. A method as recited in claim 1, wherein determining said geometric descriptors comprises:

determining an area of said boundary of said feature; and

determining an average density of selected portions of said feature; and

wherein determining whether said abnormality exists comprises;

comparing said area of said boundary of said feature to an area determined for a normal feature; and

comparing said average density for each of said portions of said feature to corresponding portions of a normal feature.

15. A method as recited in claim 14, wherein comparing said area and comparing said average density each comprises performing a regression line analysis.

7. A method as recited in 6, wherein performing local gray-level thresholding comprises:

placing regions-of-interest having a plurality of pixels along said initial boundary;

determining a second gray-level threshold value for each of said regions-of-interest;

selecting pixels in said regions-of-interest based upon respective ones of said second gray-level threshold values to obtain selected pixels; and

forming a second binary image based upon said selected pixels.

8. A method as recited in claim 7, comprising:

determining whether a first region-of-interest among said regions-of-interest is located on a first predetermined portion of said initial boundary;

constructing a second gray-level histogram for pixels located within said first region-of-interest determined to be located on said predetermined portion of said initial boundary;

determining said second gray-level threshold value for said first region-of-interest using said second gray-level histogram;

determining whether a second region-of-interest among said regions-of-interest is located on a second predetermined portion of said initial boundary; and

determining said second gray-level threshold value for said second region-of interest determined to be located on said second predetermined portion of said initial boundary based upon gray-level values of pixels located within said second region-of interest.

9. A method as recited in claim 8, wherein determining said second gray-level threshold value for said first region-of-interest comprises:

locating at least one minimum point in said second gray-level histogram; and

selecting said second gray-level threshold value for said first region-of-interest as a gray-level value corresponding to said minimum point.

10. A method as recited in claim 9, comprising:

selecting said second gray-level threshold value for said first region-of-interest as a gray-level value corresponding to a minimum point in said second gray-level histogram having a greatest gray-level value.

11. A method as recited in claim 10, comprising:

selecting a third gray-level threshold value for said first region-of-interest as an average of said second gray-level threshold value for said first region-of-interest and gray-level threshold values for two regions-of-interest neighboring said first region-of-interest.

12. A method as recited in claim 7, comprising:

detecting a boundary of said second binary image;

performing a morphological open operation on said boundary; and

superposing said boundary on said radiographic image.

13. A method as recited in claim 7, comprising:

detecting a boundary of said second binary image;

determining said geometric descriptors using said boundary of said second binary image;

detecting whether said abnormality exists using said geometric descriptors calculated using said boundary of said second binary image.

16. A method for detecting an abnormality in a radiographic chest image, comprising:

obtaining a radiographic chest image of a subject;

detecting a boundary of lungs in said radiographic image;

determining geometric descriptors of said lungs;

determining whether an abnormality exists in said lungs using said geometric descriptors;

selectively positioning a first region-of-interest over the thorax in said radiographic chest image;

constructing a gray-level histogram using pixels in said first region-of-interest;

performing a running slope of said gray-level histogram to construct a smoothed gray-level histogram; and

determining a number of first gray-level threshold values at selected locations in said smoothed gray-level histogram.

17. A method as recited in claim 16, comprising:

obtaining a radiographic chest image having a plurality of pixels;

determining horizontal gray-level profiles in said radiographic image; and

determining lung apices and mediastinum locations using said horizontal gray-level profiles.

18. A method as recited in claim 16, further comprising:

creating respective binary images at each of said number of first gray-level threshold values; and

detecting a region corresponding to said lungs using said binary images.

19. A method as recited in claim 18, comprising:

using said binary images to create an initial boundary of said lungs; and

performing local gray-level thresholding using said initial boundary of said lungs.

20. A method as recited in 19, wherein performing local gray-level thresholding comprises:

placing regions-of-interest having a plurality of pixels along said initial boundary;

determining a second gray-level threshold value for each said region-of-interest;

selecting pixels in said regions-of-interest based upon respective ones of said second gray-level threshold values to obtain selected pixels; and

forming a second binary image based upon said selected pixels.

21. A method as recited in claim 20, comprising:

placing a region-of-interest over a costophrenic angle of said lungs; and

placing a region-of-interest over the cardiophrenic angle of said lungs.

22. A method as recited in claim 20, comprising:

determining whether a first region-of-interest among said region-of-interests is located on a lateral aspect of said lung;

constructing a second gray-level histogram for pixels located within said first region-of-interest determined to be located on said lateral aspect of said lung;

determining said second gray-level threshold value for said first region-of-interest using said second gray-level histogram;

determining whether a second region-of-interest among said region-of-interests is located on at least one of a medial and diaphragmatic aspects of said lungs; and

determining said second gray-level threshold value for said second region-of-interest determined to be located on said second predetermined portion of said initial boundary based upon gray-level values of pixels located within said second region-of-interest.

23. A method as recited in claim 22, wherein determining said second gray-level threshold value for said first region-of-interest comprises:

locating a minimum in said second gray-level histogram; and

selecting said second gray-level threshold value for said first region-of-interest as a gray-value corresponding to said minimum.

24. A method as recited in claim 23, comprising:

selecting said second gray-level threshold value for said first region-of-interest as a gray-level value corresponding to a minimum in said second gray-level histogram having a greatest gray-level value.

25. A method as recited in claim 20, comprising:

detecting a boundary of said second binary image;

determining said geometric descriptors using said boundary of said second binary image;

detecting whether said abnormality exists using said geometric descriptors calculated using said boundary of said second binary image.

26. A method as recited in claim 25, wherein determining said geometric descriptors comprises:

detecting right and left lung boundaries;

determining an area of said right lung boundary and an area of said left lung boundary;

dividing said each of said right and left lung boundaries into a selected number of portions; and

determining an average density of pixels in each of said portions of said right and left lung boundaries;

and wherein determining whether said abnormality exists comprises;

determining a ratio of said area of said right lung boundary to said area of said left lung boundary;

comparing said ratio to a corresponding ratio determined for normal lungs; and

comparing said average density for each of said portions of said right and left lung boundaries to a corresponding average densities of normal lungs.

27. A method as recited in claim 26, wherein comparing said ratio and comparing said average density each comprises performing a regression line analysis.

28. A method as recited in claim 27, comprising:

determining a regression line based upon said areas of said left and right lung boundaries and upon areas for normal left and right lungs;

determining a threshold of a distance for a point from said regression line using said areas for normal left and right lungs; and

determining whether said abnormality exists based upon said threshold.

29. A method as recited in claim 27, comprising:

determining a regression line based upon said densities of said portions of said left and right lung boundaries and upon said corresponding densities of said normal lungs; and

determining a threshold of a distance for a point from said regression line using said corresponding densities of said normal lungs; and

determining whether said abnormality exists based upon said threshold.

30. A method as recited in claim 27, comprising:

determining a first regression line based upon said areas of said left and right lung boundaries and upon areas for normal left and right lungs;

determining a first threshold of a distance for a point from said regression line using said areas for normal left and right lungs;

determining a second regression line based upon said densities of said portions of said left and right lung boundaries and upon said corresponding densities of said normal lungs; and

determining a second threshold of a distance for a point from said regression line using said corresponding densities of said normal lungs; and

determining whether said abnormality exists based upon at least one of said first and second thresholds.

31. A system for detecting an abnormality in a radiographic image, comprising:

an image acquisition device;

an image memory connected to said image acquisition device;

an iterative global thresholding circuit connected to said image memory;

a centroid circuit connected to said iterative global threshold circuit;

a local thresholding circuit connected to said centroid circuit;

a correction circuit connected to said local thresholding circuit;

a sizing circuit connected to said correction circuit;

a density circuit connected to said correction circuit;

a comparison circuit connected to said sizing and density circuits; and

a display connected to receive an input of said comparison circuit.

32. A system as recited in claim 31, wherein said iterative global thresholding circuit comprises:

means for constructing a gray-level histogram of said radiographic image;

means for performing a running slope of said gray-level histogram to obtain a smoothed gray-level histogram; and

means for determining a number of first gray-level threshold values at selected locations in said smoothed gray-level histogram.

33. A system as recited in claim 31, wherein said iterative global thresholding circuit further comprises:

means for creating respective binary images at each of said number of first gray-level threshold values;

means for detecting a region corresponding to a feature in said radiographic image using said binary images; and

means for using said binary images to create an initial boundary of said feature.

34. A system as recited in claim 33, wherein said local thresholding circuit comprises:

means for placing regions-of interest having a plurality of pixels along said initial boundary;

means for determining a second gray-level threshold value for each said region-of-interest;

means for selecting pixels in said regions-of-interest based upon respective ones of said second gray-level threshold values to obtain selected pixels; and

means for forming a second binary image based upon said selected pixels.

35. A system as recited in claim 34, wherein said local thresholding circuit further comprises:

means for determining whether a first region-of-interest among said regions-of-interest is located on a first predetermined portion of said initial boundary;

means for constructing a second gray-level histogram for pixels located within said first region-of-interest determined to be located on said predetermined portion of said initial boundary;

means for determining said second gray-level threshold value for said first region-of-interest using said second gray-level histogram;

means for determining whether a second region-of-interest among said regions-of-interest is located on a second predetermined portion of said initial boundary; and

means for determining said second gray-level threshold value for said second region-of-interest determined to be located on said second predetermined portion of said initial boundary based upon gray-level values of pixels located within said second region-of-interest.

36. A system as recited in claim 31, wherein:

said sizing circuit comprises means for determining an area of a boundary of said feature;

said density circuit comprises means for determining an average density of selected portions of said feature; and

said comparison circuit comprises;

means for comparing said area of said boundary of said feature to an area determined for a normal feature; and

means for comparing said average density for each of said portions of said feature to corresponding portions of a normal feature.

37. A system as recited in claim 36, wherein comparison circuit further comprises means for performing a regression line analysis.

38. A method for automated analysis of a radiographic image, comprising:

obtaining a radiographic image of a subject; and

detecting a boundary of a feature in said radiographic image by performing iterative global gray-level thresholding on said radiographic image, comprising,

determining a plurality of gray-level threshold values

iteratively creating respective binary images defining at least one region based on each of said plurality of said gray-level threshold values,

defining a characteristic limit for said at least one region,

eliminating said at least one region from the respective image and subsequent images iteratively obtained if said at least one region has a characteristic outside said characteristic limit, and

determining a final region resulting after iterative eliminating of regions and iterative thresholding at said plurality of gray level threshold values as said feature.

39. A method as recited in claim 38, wherein said step of iteratively creating said respective binary images comprises turning on pixels in each iterative image having a gray level less than a respective one of said gray-level threshold values.

40. A method as recited in claim 38, comprising:

said step of defining a characteristic limit comprising defining a centroid limit in said radiographic image; and

said step of eliminating comprising eliminating said region if it has a centroid outside said centroid limit.

41. A method as recited in claim 38, comprising:

performing a morphological open operation on selected ones of said binary images.

42. A method as recited in claim 38, comprising:

using said binary images to create an initial boundary of said feature; and

performing local gray-level thresholding using said initial boundary of said feature.

43. A method as recited in claim 42, wherein performing local gray-level thresholding comprises:

placing regions-of interest having a plurality of pixels on said initial boundary;

determining a second gay-level threshold value for each said region-of-interest;

selecting pixels in said regions-of interest based upon respective ones of said second gray-level threshold values to obtain selected pixels; and

forming a second binary image based upon said selected pixels.

44. A method as recited in claim 43, comprising:

determining whether a first region-of-interest among said region-of-interests is located on a first predetermined portion of said initial boundary;

constructing a second gray-level histograms for pixels located within said first region-of interest determined to be located on said predetermined portion of said initial boundary;

determining said second gray-level threshold value for said first region-of-interest using said second gray-level histogram;

determining whether a second region-of-interest among said region-of-interests is located on a second predetermined portion of said initial boundary; and

determining said secondary gray-level threshold value for said second region-of-interest determined to be located on said second predetermined portion of said initial boundary based upon gray-level values of pixels located within said second region-of interest.

45. A method as recited in claim 44, wherein determining said second gray-level threshold value for said first region-of-interest comprises:

locating at least one minimum point in said second gray-level histogram; and

selecting said second gray-level threshold value for said first region-of-interest as a gray-value corresponding to said minimum.

46. A method as recited in claim 45, comprising:

selecting said second gray-level threshold value for said first region-of-interest as a gray-level value corresponding to a minimum in said second gray-level histogram having a greatest gray-level value.

47. A method as recited in claim 46, comprising:

selecting a third gray-level threshold value for said first region-of-interest as an average of said second gray-level threshold value for said first region-of interest and second gray-level values for two regions-of-interest neighboring said first region-of-interest.

48. A method as recited in claim 43, comprising:

detecting a boundary of said second binary image;

performing a morphological open operation on said boundary; and

superposing said boundary on said radiographic image.