Automated method and system for the detection of lesions in medical computed tomographic scans

• US 5,881,124 A
• Filed: 03/31/1994
• Issued: 03/09/1999
• Est. Priority Date: 03/31/1994
• Status: Expired due to Term

• Current Assignee: ARCH Development Corporation (University of Chicago)
• Original Assignee: ARCH Development Corporation (University of Chicago)
• Inventors: Bae, Kyongtae Ty, Giger, Maryellen L., Doi, Kunio
• Primary Examiner(s): Porta, David P.
• Assistant Examiner(s): Bruce, David Vernon
• Application Number:
  US08/220,917
• Time in Patent Office: 1,804 Days
• Field of Search: 378/4, 378/8, 378/901, 364/413.14, 382/131, 250/363.04
• US Class Current: 378/8
• CPC Class Codes: G06T 2207/10081 Computed x-ray tomography [CT]
• CPC Class Codes: G06T 2207/30064 Lung nodule
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: Y10S 378/901 Computer tomography program or processor

Abstract

A method and system for the automated detection of lesions in computed tomographic images, including generating image data from at least one selected portion of an object, for example, from CT images of the thorax. The image data are then analyzed in order to produce the boundary of the thorax. The image data within the thoracic boundary is then further analyzed to produce boundaries of the lung regions using predetermined criteria. Features within the lung regions are then extracted using multi-gray-level thresholding and correlation between resulting multi-level threshold images and between at least adjacent sections. Classification of the features as abnormal lesions or normal anatomic features is then performed using geometric features yielding a likelihood of being an abnormal lesion along with its location in either the 2-D image section or in the 3-D space of the object.

Claims

1. A method for the automated detection of nodules in a computed tomographic (CT) scan of a subject, comprising:

detecting an anatomic region of said subject in said CT scan; and

detecting a nodule in said anatomic region using a plurality of images of said anatomic region in said CT scan generated at different section-specific gray-level thresholds.

2. A method as recited in claim 1, wherein said step of detecting said anatomic region comprises:

forming a histogram of gray-values of pixels in said CT scan;

determining a gray-level threshold using said histogram; and

generating a binary image of said anatomic region using said gray level threshold.

3. A method as recited in claim 2, comprising:

forming a histogram of gray-values of pixels in said CT scan from a corner of said CT scan to a center of said scan;

wherein said step of selecting said gray level threshold comprises;

determining a first cumulative mean of said histogram from said corner to said center of said CT scan;

determining a second cumulative mean of said histogram from said center to said corner of said CT scan;

determining a difference between said first and second cumulative means; and

using said difference to determine said gray-level threshold.

4. A method as recited in claim 2, wherein said step of detecting said anatomic region further comprises:

detecting a boundary of said binary image; and

evaluating geometric parameters of a region enclosed by said boundary.

5. A method as recited in claim 4, wherein said step of detecting a boundary comprises detecting a thoracic boundary.

6. A method as recited in claim 4, wherein:

said step of detecting said boundary of said binary image comprises using 8-point connectivity; and

said step of evaluating geometric parameters comprises evaluating a size and a compactness of said region.

7. A method as recited in claim 6, wherein said step of evaluating geometric parameters further comprises:

requiring said size of said region to be at least a predetermined portion of a size of said CT scan and requiring said compactness of said region to be at least a predetermined value in order for said region to be detected as said anatomic region.

8. A method as recited in claim 2, comprising:

detecting a boundary in said binary image;

forming a second gray-level histogram of pixels within said boundary;

determining a gray-level threshold based on said histogram;

generating a second binary image using said second gray-level threshold.

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

detecting a second boundary in said second binary image; and

evaluating geometric parameters of a region enclosed by said second boundary.

10. A method as recited in claim 9, wherein said step of evaluating geometric parameters comprises:

determining an area of said region; and

correcting said area using a corner correction technique.

11. A method as recited in claim 9, wherein said step of evaluating geometric parameters comprises:

requiring said region to be located in said first boundary and requiring a compactness of said region to be at least a predetermined value in order for said region to be detected as said anatomic region.

12. A method as recited in claim 11, wherein:

detecting said first boundary comprises detecting a thoracic boundary; and

said region is a lung.

13. A method as recited in claim 1, further comprising:

detecting features in said anatomic region as suspected nodules using said plurality of images;

determining geometric descriptors for selected ones of said features; and

detecting said nodule using said geometric descriptors.

14. A method as recited in claim 13, wherein said step of determining geometric descriptors comprises:

determining plural of perimeter, area, compactness, elongation, circularity, distance measure and total score for said selected ones of said features.

15. A method as recited in claim 13, wherein said step of detecting said nodule using said geometric descriptors comprises assigning each of said features a likelihood of being a nodule.

16. A method as recited in claim 15, wherein:

said anatomic region is a lung pair; and

said step of detecting said nodule using said geometric descriptors comprises assigning each of said features a likelihood of being a nodule or a vessel.

17. A method as recited in claim 1, further comprising:

detecting features in said anatomic region as suspected nodules using said plurality of images; and

performing a corner correction routine on selected ones of said features.

18. A method as recited in claim 17, wherein performing said corner correction routine comprises:

detecting a boundary of each of said selected ones of said features;

determining chain codes for pixels on said boundary of each of said selected ones of said features; and

selectively reducing or expanding an area of said pixels on said border based upon said chain codes.

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

determining chain codes B1-B7 for a possible next boundary pixel for each pixel A on said boundary of said feature;

selectively reducing or expanding an area of each pixel A on said boundary according to said chain codes B1-B7 as;

B1--reduce area of pixel by 3/8

B2--reduce area of pixel by 2/8

B3--reduce area of pixel by 1/8

B4--no reduction or expansion

B5--expand area of pixel by 1/8

B6--expand area of pixel by 2/8

B7--expand area of pixel by 3/8; and

generating a corner correction term based on said chain codes.

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

said chain codes B1-B7 having direction values of B1=5, B2=6, B3=7, B4=0, B5=1, B6=2 and B7=3;

said pixel A having a direction value of 8; and

determining Σ

(Bi-A) over all pixels on said boundary as said corner correction term.

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

determining features in anatomic region using said images; and

forming a tree structure of features in said binary images.

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

generating said images using a corresponding plurality of successively stricter gray-level thresholds;

wherein said step of forming tree structure comprises;

selecting one of said images;

detecting a first node within said one image;

selecting a second one of said images generated at a next stricter gray-level threshold;

detecting a second node within said second image; and

determining if a position of said second node lies within a corresponding position of said first node in said second image.

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

determining a boundary of said first node; and

determining if a center of said second node lies within a corresponding boundary of said first node in said second image.

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

said CT scan comprising a plurality of CT sections; and

generating said plurality of images for each of said plurality of CT sections.

25. A method as recited in claim 1, wherein:

said step of detecting a nodule within said anatomic region comprises generating a plurality of binary images of said anatomic region at a corresponding plurality of different gray-level thresholds;

said method comprises;

forming a histogram of gray values of pixels in said anatomic region;

generating said plurality of binary images of said anatomic region at gray-level threshold values corresponding to pixel values at predetermined portions of an area under said histogram;

detecting features in said anatomic region as suspected nodules; and

detecting a boundary of each feature in each of said binary images.

26. A method as recited in claim 1, wherein:

said CT scan comprises a plurality of CT sections;

said step of detecting a nodule comprises generating a plurality of images for each of said CT sections;

said step of detecting a nodule comprises;

determining features in anatomic region using said images;

determining geometric descriptors for selected ones of said features; and

using a rule-based system having rules based on said geometric descriptors to assign said selected ones of said features a likelihood of being a nodule; and

said method comprises using said rule-based system having rules determined from cumulative frequencies of said features being a nodule.

27. A method as recited in claim 26, comprising:

forming a tree structure defining as nodes said features detected in said plurality of images of said anatomic region;

determining plural of perimeter, area compactness, elongation, circularity, distance measure and total score as said geometric descriptors;

determining a boundary for said selected ones of said features;

wherein said step of using a rule-based system comprises;

determining whether one of said nodes has a corresponding node in an adjacent one of said plurality of images and, if said corresponding node is present in said adjacent one image deleting said one node and evaluating geometric descriptors of said corresponding node;

evaluating geometric descriptors of said one node if said corresponding node is not present.

28. A method as recited in claim 27, wherein said step evaluating geometric descriptors of said one node comprises determining whether said one node is said anatomic region.

29. A method as recited in claim 26, further comprising generating a analyzed CT scan with identified features in said anatomic region.

30. A method as recited in claim 27, wherein said CT scan comprises a plurality of CT sections;

and

said method further comprises generating a plurality of analyzed CT sections each having identified features in said anatomic region.

31. A method as recited in claim 30, wherein said step of comparing comprises:

assigning each of said nodes a likelihood of being a nodule;

determining which of said analyzed CT sections contain nodes that have not been definitely identified;

comparing each of said analyzed CT sections having nodes that have not been definitely identified with adjacent ones of said analyzed CT sections; and

updating a likelihood of being a nodule for each node having a corresponding node in said adjacent ones of said analyzed CT sections to a likelihood of being a nodule of said corresponding node when said corresponding node has a more definite likelihood of being a nodule.

32. A method as recited in claim 31, wherein said step of comparing each of said analyzed CT sections comprises:

determining locations of all nodes in one analyzed CT section assigned a likelihood of being a normal anatomic feature;

determining nodes having corresponding locations in an analyzed CT section adjacent to said one analyzed CT section which overlap said locations each of said nodes in said one analyzed CT section;

updating a likelihood of being a nodule for each of said nodes in said one analyzed CT section based upon corresponding overlapping nodes.

33. A method as recited in claim 32, wherein said step of updating a likelihood comprises:

determining a first count of pixels of a first node in said one analyzed section;

determining a second count of pixels of nodes in adjacent analyzed CT sections determined to overlap said first node;

updating a likelihood of said first node based upon said first and second counts.

34. A system for the automated detection of lesions in computed tomographic (CT) scan, comprising;

an image generation device to generate said CT scan of an anatomic region;

a device for generating gray-level images of said CT scan at a plurality of gray-level thresholds; and

a device for analyzing features in said gray-level images;

wherein said device for generating gray-level threshold images comprises;

a histogram generator which generates a histogram of gray-values of pixels in said CT scan;

a threshold generator for determining a plurality of gray-level thresholds using said histogram; and

a binary image generator for generating a respective plurality of binary images of said anatomic region using said gray-level thresholds;

said CT scan comprising a plurality of CT sections;

said binary image generator generating a respective plurality of binary images for each of said CT sections; and

a tree structure generator producing a tree structure of features in said respective plurality of binary images for each CT section.

35. A system as recited in claim 34, comprising:

a feature detector for detecting features using said binary image of said anatomic region;

a geometric parameter generator to generate geometric parameters of said features; and

a rule-based feature analyzer which analyzes said features using said geometric parameters.

36. A system as recited in claim 35, comprising:

said CT scan comprising a plurality of CT sections;

said rule-based feature analyzer generating a plurality of analyzed CT sections having features assigned a status indicating a likelihood of being a nodule;

a section comparer for comparing said plurality of CT sections and updating a status of a feature in one section based upon a status of a feature in an adjacent section.

37. A method for the automated detection of nodules in a computed tomographic (CT) scan of a subject, comprising:

detecting an anatomic region of said subject in said CT scan;

detecting a nodule in said anatomic region using a plurality of images of said anatomic region in said CT scan generated at different gray-level thresholds;

detecting features in said anatomic region as suspected nodules using said plurality of images;

determining geometric descriptors for selected ones of said features; and

detecting said nodule using said geometric descriptors;

wherein said step of determining geometric descriptors comprises determining plural of perimeter, area, compactness, elongation, circularity, distance measure and total score for said selected ones of said features.

38. A method for the automated detection of nodules in a computed tomographic (CT) scan of a subject, comprising:

detecting an anatomic region of said subject in said CT scan;

detecting a nodule in said anatomic region using a plurality of images of said anatomic region in said CT scan generated at different gray-level thresholds;

detecting features in said anatomic region as suspected nodules using said plurality of images;

determining geometric descriptors for selected ones of said features; and

detecting said nodule using said geometric descriptors;

wherein said step of detecting said nodule using said geometric descriptors comprises assigning each of said features a likelihood of being a nodule.

39. A method for the automated detection of nodules in a computed tomographic (CT) scan of a subject, comprising:

detecting an anatomic region of said subject in said CT scan;

detecting a nodule in said anatomic region using a plurality of images of said anatomic region in said CT scan generated at different gray-level thresholds;

detecting features in said anatomic region as suspected nodules using said plurality of images; and

performing a corner correction routine on selected ones of said features.

40. A method for the automated detection of nodules in a computed tomographic (CT) scan of a subject, comprising:

detecting an anatomic region of said subject in said CT scan;

detecting a nodule in said anatomic region using a plurality of images of said anatomic region in said CT scan generated at different gray-level thresholds;

determining features in anatomic region using said images; and

forming a tree structure of features in said binary images.

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

generating said images using a corresponding plurality of successively stricter gray-level thresholds;

wherein said step of forming tree structure comprises;

selecting one of said images;

detecting a first node within said one image;

selecting a second one of said images generated at a next stricter gray-level threshold;

detecting a second node within said second image; and

determining if a position of said second node lies within a corresponding position of said first node in said second image.

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

determining a boundary of said first node; and

determining if a center of said second node lies within a corresponding boundary of said first node in said second image.

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

said CT scan comprising a plurality of CT sections; and

generating said plurality of images for each of said plurality of CT sections.