Method and system for automatic detection of ribs and pneumothorax in digital chest radiographs

• US 5,668,888 A
• Filed: 12/29/1993
• Issued: 09/16/1997
• Est. Priority Date: 11/21/1990
• Status: Expired due to Term

• Current Assignee: ARCH Development Corporation (University of Chicago)
• Original Assignee: ARCH Development Corporation (University of Chicago)
• Inventors: Sanada, Shigeru, Doi, Kunio
• Primary Examiner(s): Weinhardt, Robert A.
• Assistant Examiner(s): Thomas, Joseph
• Application Number:
  US08/174,175
• Time in Patent Office: 1,357 Days
• Field of Search: 364/413.13, 382/6, 382/9, 382/16, 382/18, 382/22, 382/54, 382/132, 382/199
• US Class Current: 382/132
• CPC Class Codes: G06T 2207/10116 X-ray image
• CPC Class Codes: G06T 2207/20116 Active contour; Active surface; Snakes
• CPC Class Codes: G06T 2207/20224 Image subtraction
• CPC Class Codes: G06T 2207/30008 Bone
• CPC Class Codes: G06T 2207/30061 Lung
• CPC Class Codes: G06T 5/50 using two or more images, e.g. averaging or subtraction
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/149 involving deformable models, e.g. active contour models
• CPC Class Codes: G06T 7/174 involving the use of two or more images

Abstract

A method and system for the detection of anatomical features in a digital chest radiograph, and in particular rib edges and pneumothorax, wherein vertical profiles are defined in regions of the lungs in the radiograph, and a model function is fitted to the profiles to obtain initial estimates of the locations of rib edges. Gradient-histogram analysis is performed by determining edge gradient and their corresponding orientations in regions of interest (ROIs) defined in the radiograph in the vicinity of the initially estimated rib edges, determining the maximum edge gradient in each of the ROIs, and forming histograms which define accurate detections of rib edges. The edges are fitted with an elliptical function and a representation of the entire rib cage structure is obtained. The method and system are also applied to the detection of pneumothorax where ROIs are defined in the apical lung region and edge gradients and their orientations are detected in the ROIs. The ROI is edge-enhanced using the edge gradients and their orientations and previously detected rib edges are removed. Subtle, curved-line structure corresponding to pneumothorax in the edge-enhanced ROI are detected and a representation of the pneumothorax is obtained using a Hough transform.

Claims

1. A method for the detection of a anatomic feature in a chest radiograph, comprising:

(a) defining plural regions of interest in said radiograph;

(b) determining edge gradients in at least one of said plural regions of interest and corresponding orientations of said edge gradients for each of a plurality of pixels included in said at least one of said regions of interest;

(c) determining a cumulative edge gradient of the edge gradients and corresponding orientations determined in said at least one of said plural regions of interest in step (b) by summing in said at least one of said plural regions of interest the edge gradients having the same orientation; and

(d) detecting at least one edge of said anatomic feature based upon the cumulative edge gradient determined in said at least one of said plural regions of interest, comprising,

(d1) determining which regions of interest in aid cumulative edge gradient have an orientation within a predetermined range of orientations; and

(d2) determining a location of an edge of said anatomic feature based on locations of regions of interest which are determined in step (d1) to have said orientation within said predetermined range of orientations; and

(e) producing a representation of said anatomic feature using said at least one detected edge.

2. A method as recited in claim 1, wherein said step (b) of determining said edge gradients comprises:

fitting said pixels in said regions of interest using a predetermined function;

subtracting said predetermined function from said pixels to obtain background-trend-corrected regions of interest; and

determining said edge gradients using pixels in said background-trend-corrected regions of interest.

3. A system according to claim 1, wherein:

step (b) comprises determining edge gradients and corresponding orientations of said edge gradients for each of the plurality of pixels included in each of said plural regions of interest;

step (c) comprises determining a respective cumulative edge gradient for each of said plural regions of interest; and

step (d1) comprises determining which of said cumulative edge gradients has a largest sum within said predetermined range of orientations; and

step (d2) comprises determining said edge of said anatomic feature as a location of the region of interest having in its respective cumulative edge gradient said largest sum of orientations within said predetermined range of orientations.

4. A method as recited in claim 3, wherein rib edges are detected, a vertical direction in said chest radiograph is a direction generally parallel to a subject'"'"'s spine and a horizontal direction is a direction perpendicular to said vertical direction, and said step (a) of defining regions of interest comprises:

determining vertical profiles in regions of both lungs of said chest radiograph;

fitting a predetermined model function to said vertical profiles to obtain initial estimates of said rib edges in said radiograph; and

defining said regions of interest in said radiograph to include regions containing said initial estimates.

5. A method as recited in claim 4, further comprising:

determining a magnitude of each of said determined edge gradients in each of said regions of interest;

determining corresponding orientations of said magnitudes of said edge gradients in a polar coordinate system having r and θ

as variables;

determining said cumulative edge gradients in each region of interest by summing said magnitudes of said edge gradients having the same edge gradient orientation in the respective region of interest; and

forming histograms of said cumulative edge gradients and using said histograms to define upper and lower edges of a rib.

6. A method as recited in claim 5, wherein said step of defining said regions of interest comprises:

defining a plurality of said regions of interest in a vertical orientation with respect to said radiograph and with respect to each other; and

overlapping each of said plurality of regions of interest by no more than half of an adjacent one of said plurality of regions of interest.

7. A method as recited in claim 4, wherein said step of producing a representation comprises:

fitting each of said detected rib edges with an elliptical function; and

obtaining a representation of an entire rib structure using said rib edges fitted with said elliptical function.

8. A method as recited in claim 7, wherein said step of fitting each of said rib edges further comprises:

grouping detected rib edge points of a selected rib based upon orientation of said edge gradients of said selected rib for distinction of upper and lower edges of said selected rib, and upon vertical positions of said detected rib edge points of said selected rib as a function of horizontal position in said radiograph.

9. A method as recited in claim 7, further comprising:

estimating a set of fitting parameters for a missing of said detected rib edges to obtain an estimated rib edge based upon previously fitted rib edges; and

reperforming said steps of defining regions of interest, determining edge gradients and detecting rib edges.

10. A system for the detection of an anatomic feature in a chest radiograph, comprising:

first means for defining regions of interest in said radiograph;

second means for determining edge gradients in said regions of interest and corresponding orientations of said edge gradients for each of a plurality of pixels included in each of said regions of interest;

third means for determining a respective cumulative edge gradient for each region of interest by summing in each region of interest the edge gradients having the same orientation;

fourth means for detecting at least one edge of said anatomic feature based upon the cumulative edge gradients determined in said plural regions of interest, comprising,

means for determining which of said cumulative edge gradients has a largest sum within a predetermined range of orientations; and

means for determining an edge of said anatomic region based on a location of the region of interest having in its respective cumulative edge gradient said largest sum within said predetermined range of orientations; and

fifth means for producing a representation of said anatomic feature using said at least one detected edge.

11. A system as recited in claim 10, wherein said second means for determining edge gradients comprises:

means for fitting said pixels in said regions of interest using a predetermined function;

means for subtracting said predetermined function from said pixels to obtain background-trend-corrected regions of interest; and

means for determining said edge gradients using pixels in said background-trend-corrected regions of interest.

12. A system as recited in claim 10, wherein said anatomic feature includes rib edges, a vertical direction in said chest radiograph is a direction generally parallel to the subject'"'"'s spine and a horizontal direction is a direction perpendicular to said vertical direction, and said first means comprises:

means for determining vertical profiles in both lung regions of said chest radiograph;

means for fitting a predetermined model function to said vertical profiles to obtain initial estimates of said rib edges in said radiograph; and

means for defining said regions of interest in said radiograph to include regions containing said initial estimates.

13. A system as recited in claim 12, wherein said fifth means comprises:

means for fitting each of said detected rib edges with an elliptical function; and

means for obtaining a representation of entire ribs using said rib edges fitted with said elliptical function.

14. A system as recited in claim 12, wherein said third means comprises:

means for determining a magnitude of each of said determined edge gradients in each of said regions of interest;

means for plotting said magnitudes of said edge gradients and corresponding orientations of said magnitudes of said edge gradients in a polar coordinate system having r and θ

as variables, wherein said magnitudes of said edge gradients are plotted as said r variable and said corresponding orientations of said edge gradient magnitudes are plotted as said θ

variable;

summing means for summing in each region of interest the magnitudes of said edge gradients having the same orientation to produce said cumulative edge gradients; and

means for forming histograms of said cumulative edge gradients and using said histograms to define upper and lower edges of a rib.

15. A system as recited in claim 14, wherein said first means for defining said regions of interest comprises:

means for defining a plurality of said regions of interest in a vertical orientation; and

means for overlapping each of said plurality of regions of interest by no more than half of an adjacent of said plurality of regions of interest.

16. A system as recited in claim 13, wherein said means for fitting each of said rib edges further comprises:

means for grouping detected rib edge points of a selected rib based upon orientation of said edge gradients of said selected rib for distinction of upper and lower edges of said selected rib, and upon vertical positions of said detected rib edge points of said selected rib as a function of horizontal position in said radiograph.

17. A system as recited in claim 13, further comprising:

means for estimating a set of fitting parameters for a missing rib edge to obtain an estimated rib edge based upon previously fitted rib edges.

18. A method for detection of a pneumothorax in a chest radiograph, comprising:

detecting rib edges in said radiograph;

defining a pneumothorax region of interest in said radiograph which includes said pneumothorax;

removing said rib edges and respective areas adjacent said rib edges from said pneumothorax region of interest;

detecting a curved line in said pneumothorax region of interest having said rib edges removed; and

producing a representation of said pneumothorax based on the detected curved line;

wherein said step of detecting rib edges comprises;

defining plural regions of interest in said radiograph which include regions containing rib edges;

determining edge gradients and corresponding orientations of the determined edge gradients for each of a plurality of pixels included in each of said regions of interest;

determining a respective cumulative edge gradient for each region of interest by summing in each region of interest the edge gradients of those pixels of the respective region of interest having the same orientation;

determining a rib edge based on the cumulative edge gradients determined in the preceding step, comprising,

determining which of said cumulative edge gradients has a largest sum within a predetermined range of angles; and

determining said rib edge based on a location of the region of interest having in its respective cumulative edge gradient said largest sum within said predetermined range of orientations.

19. A method as recited in claim 18, wherein a vertical direction in said chest radiograph is a direction generally parallel to a subject'"'"'s spine and a horizontal direction is a direction perpendicular to said vertical direction, and said step of defining said pneumothorax region of interest comprises:

determining vertical profiles in regions of both lungs of said chest radiograph;

fitting a predetermined model function to said vertical profiles to obtain initial estimates of said rib edges in said radiograph; and

defining first regions of interest in said radiograph which include regions containing said initial estimates.

20. A method as recited in claim 18, wherein said step of removing said rib edges comprises:

fitting said detected rib edges with a predetermined function providing a rib edge curve for each of said detected rib edges; and

removing from said pneumothorax region of interest first pixels corresponding to said rib edge curves and a predetermined number of second pixels surrounding each of said first pixels on each of said rib edge curves.

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

edge enhancing said pneumothorax region of interest by determining edge gradients and corresponding orientations for each pixel in said pneumothorax region of interest and assigning data values to pixels having edge gradients at predetermined orientations and exceeding predetermined thresholds;

defining islands in said edge-enhanced pneumothorax region of interest;

determining a maximum magnitude of said edge gradients and a sum of magnitudes of said edge gradients in each island in said edge-enhanced pneumothorax region of interest;

eliminating first islands having a sum of magnitudes of said edge gradients below a first predetermined value in said edge-enhanced pneumothorax region of interest; and

eliminating second islands having a maximum of magnitudes of said edge gradients above a second predetermined value in said edge-enhanced pneumothorax region of interest.

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

said step of eliminating said first islands comprises removing noise components from said edge-enhanced pneumothorax region of interest; and

said step of eliminating said second islands comprises removing remaining rib edge components from said edge-enhanced pneumothorax region of interest.

23. A method as recited in claim 18, wherein said step of detecting said curved line comprises:

using a Hough transform to detect said curved line corresponding to said pneumothorax;

obtaining a strongly accumulated area having accumulated values in a r-θ

parameter space using said Hough transform; and

detecting said pneumothorax by applying an inverse Hough transform to selected points in said strongly accumulated area in said r-θ

parameter space.

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

eliminating in said strongly accumulated values a first portion of said accumulated values below a first predetermined threshold; and

eliminating in said strongly accumulated area a second portion of said accumulated values above a second predetermined threshold.

25. A system as recited in claim 24, wherein:

said step of eliminating said first portion of said accumulated values comprises removing noise components from said strongly accumulated area; and

said step of eliminating said second portion of said accumulated values comprises removing rib edge components from said strongly accumulated area.

26. A system for detection of a pneumothorax in a chest radiograph, comprising:

means for defining rib edge regions of interest and a pneumothorax region of interest;

first means for detecting rib edges in said rib edge regions of interest in said radiograph;

second means for removing said rib edges and respective adjacent areas from said pneumothorax region of interest;

third means for detecting a curved line in said enhanced pneumothorax region of interest having said rib edges removed; and

fourth means for producing representation of said pneumothorax based on the detected curve line;

wherein said first means comprises;

means for determining edge gradients and corresponding orientations of the determined edge gradients for each of a plurality of pixels included in each of said rib edge regions of interest;

means for determining a respective cumulative edge gradient for each rib edge region of interest by summing the edge gradients of pixels of the respective rib edge region of interest which have the same orientation;

means for determining a rib edge based on said respective cumulative edge gradients, comprising,

means for determining which of said respective cumulative edge gradients has a greatest sum within a predetermined range of orientations, and

means for determining said rib edge based on a location of the region of interest having in its respective cumulative edge gradient said largest sum in said predetermined range of orientations.

27. A system as recited in claim 26, wherein a vertical direction in said chest radiograph is a direction generally parallel to a subject'"'"'s spine, a horizontal direction is a direction perpendicular to said vertical direction, and said means for defining regions of interest comprises:

means for determining vertical profiles in both lung regions of said chest radiograph;

means for fitting a predetermined model function to said vertical profiles to obtain initial estimates of said rib edges in said radiograph; and

means for defining said regions of interest in said radiograph as regions containing said initial estimates.

28. A system as recited in claim 26, wherein said second means for removing said rib edges comprises:

means for fitting said detected rib edges with a predetermined function providing a rib edge curve for each of said detected rib edges; and

means for removing from said pneumothorax region of interest first pixels corresponding to said rib edge curves and a predetermined number of second pixels surrounding each of said first pixels on each of said rib edge curves.

29. A system as recited in claim 26, further comprising:

means for edge enhancing said pneumothorax region of interest by determining edge gradients and corresponding orientations for each pixel in said pneumothorax region of interest and assigning data values to those pixels having edge gradients at predetermined orientations and exceeding predetermined thresholds;

means for defining islands in said edge-enhanced pneumothorax region of interest;

means for determining a maximum magnitude of said edge gradients and a sum of magnitudes of said edge gradients in each island in said edge-enhanced pneumothorax region of interest;

means for eliminating first islands having a sum of magnitudes of said edge gradients below a first predetermined value in said edge-enhanced pneumothorax region of interest; and

means for eliminating second islands having a maximum of magnitudes of said edge gradients above a second predetermined value in said edge-enhanced pneumothorax region of interest.

30. A system as recited in claim 29, wherein:

said means for eliminating said first islands comprises means for removing noise components from said edge-enhanced pneumothorax region of interest; and

said means for eliminating said second islands comprises means for removing remaining rib edge components from said edge-enhanced pneumothorax region of interest.

31. A system as recited in claim 26, wherein said third means comprises:

means for using a Hough transform to detect said curved line corresponding to said pneumothorax;

means for obtaining a strongly accumulated area having accumulated values in a r-θ

parameter space using said Hough transform; and

means for detecting said pneumothorax by applying an inverse Hough transform to selected points in said strongly accumulated area in said r-θ

parameter space.

32. A system as recited in claim 31, further comprising:

means for eliminating a first portion of said accumulated values in said strongly accumulated area below a first predetermined threshold; and

means for eliminating a second portion of said accumulated values in said strongly accumulated area above a second predetermined threshold.

33. A system as recited in claim 32, wherein:

said means for eliminating said first portion of said accumulated values comprises means for removing noise components from said strongly accumulated area; and

said means for eliminating said second portion of said accumulated values comprises means for removing rib edge components from said strongly accumulated area.