Method and system for automated computerized analysis of sizes of hearts and lungs in digital chest radiographs

• US 5,072,384 A
• Filed: 11/23/1988
• Issued: 12/10/1991
• Est. Priority Date: 11/23/1988
• Status: Expired due to Term

• Current Assignee: ARCH Development Corporation (University of Chicago)
• Original Assignee: ARCH Development Corporation (University of Chicago)
• Inventors: Nakamori, Nobuyuki, Doi, Kunio
• Primary Examiner(s): Shaw, Dale M.
• Assistant Examiner(s): Huntley, David
• Application Number:
  US07/275,720
• Time in Patent Office: 1,112 Days
• Field of Search: 364/413.13, 364/413.19, 382/6, 382/9, 382/22
• US Class Current: 382/132
• CPC Class Codes: G06T 2207/10116 X-ray image
• CPC Class Codes: G06T 2207/20056 Discrete and fast Fourier transform, [DFT, FFT]
• CPC Class Codes: G06T 2207/30048 Heart; Cardiac
• CPC Class Codes: G06T 2207/30061 Lung
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/66 of image moments or centre of gravity

Abstract

An automated method and system to determine a number of parameters related to the size and shape of the heart as well as parameters related to the lungs from data derived from digital chest radiographs. A cardiac rectangle enclosing the heart and portions of the surrounding lung tissue is determined, and within the cardiac rectangle, horizontal and vertical profiles, and the first derivatives thereof, are determined. Based on these derivatives, cardiac boundary points on the left and right sides of the cardiac contour are determined, as well as diaphragm edge points. A predetermined model function is then fitted to selected of the determined cardiac boundary points to determine the cardiac contour. Tests are performed to determine whether or not the heart has an abnormal size or is a "tall" heart, and if so, corrective measures are taken. In a preferred embodiment, a shift-variant cosine function is used as a model function fitted to the selected cardiac boundary points. In an alternative embodiment, the model function is equivalent to the partial summation of a Fourier series. In an alternative embodiment for determining cardiac boundaries, an analysis is made of edge gradients obtained in two orthogonal directions in plural narrow band regions of the data from the digital chest radiograph.

Claims

1. A method for determining a cardiac contour of a subject from a chest radiograph of the subject, comprising:

detecting plural right and left cardiac boundary points in the cardiac contour;

fitting a predetermined model function to the detected cardiac boundary points to derive a completed cardiac contour based on the fitted model function;

using a shift-variant sinusoidal function as said predetermined model function; and

producing a representation of the completed cardiac contour.

2. The method according to claim 1, comprising:

selecting said shift-variant sinusoidal function f₁ (θ

) defined by the following relationship

space="preserve" listing-type="equation">f.sub.1 (θ

)=r.sub.o +r.sub.1 cos {2(θ

-φ

)-α

g(θ

-φ

)},

where r_o and r₁ are coefficients corresponding to the mean radial distance from center and the maximum variation from the mean radial distance, respectively, when the cardiac contour is expressed in a polar coordinate system and θ

is the angle represented in said polar coordinate system, and α and

φ

are computed coefficients, wherein coefficients r₀, r₁, φ and

α

are determined by means of a least squares method, and where g(θ

) is a phase term which is a predetermined function of angle θ

.

3. The method according to claim 2, comprising:

selecting the phase term g(θ

) from the following group consisting of;

space="preserve" listing-type="equation">g.sub.1 (θ

)=cos θ

,

space="preserve" listing-type="equation">g.sub.2 (θ

)=sin (θ

),

space="preserve" listing-type="equation">g.sub.3 (θ

)=1-|θ

|/π

,

space="preserve" listing-type="equation">g.sub.4 (θ

)=|θ

|/π

,

space="preserve" listing-type="equation">g.sub.5 (θ

)=1-(θ

/π

).sup.2,

and

space="preserve" listing-type="equation">g.sub.6 (θ

)=(θ

/π

).sup.2.

4. The method according to claim 1, wherein said step of detecting plural right and left cardiac boundary points comprises:

obtaining edge gradients (G_hx, G_hy) in two orthogonal directions from pixel values located in plural narrow horizontal bands of pixels in said chest radiograph; and

selecting cardiac boundary points based on the obtained edge gradients.

5. The method according to claim 4, wherein said step of selecting cardiac boundary points comprises:

determining a cardiac midline (MDL) dividing the cardiac region of the chest radiograph into right and left portions;

determining an overall edge gradient G_h and an edge gradient orientation angle θ

_h based on the obtained edge gradients (G_hx, G_hy), where ##EQU3## for at least selected portions of each narrow horizontal band of pixels;

selecting as possible cardiac boundary points on opposite sides of the cardiac midline MDL the point in each narrow horizontal band of pixels at which G_h is a maximum;

selecting as a lowest cardiac boundary point from the possible cardiac boundary points selected in the preceding step that point of the possible cardiac boundary points which exhibits a minimum peak in orientation angle θ

_h ; and

rejecting as not being cardiac boundary points those possible cardiac boundary points extending below said minimum boundary point.

6. The method according to claim 5, comprising:

detecting diaphragm edge points, including the steps of,

obtaining edge gradients (G_vx, G_vy) in two orthogonal directions from pixel values located in plural narrow vertical bands of pixels in said chest radiograph, and

selecting diaphragm edge points based on the edge gradients obtained in the preceding step.

7. The method according to claim 6, wherein said step of selecting diaphragm edge points comprises:

determining an overall edge gradient (G_v) based on the edge gradients (G_vx, G_vy), where ##EQU4## for at least selected portions of each narrow vertical band of pixels;

selecting as a prospective diaphragm edge point in each narrow vertical band of pixels that point at which G_v is maximum; and

connecting said prospective diaphragm edge points with a third order polynomial curve which defines a diaphragm edge.

8. The method according to claim 7, comprising:

selecting plural of those possible cardiac boundary points, which are adjacent said diaphragm edge points as cardiac boundary points;

fitting said selected cardiac boundary points to a predetermined model function in the form of a circle using a least squares method to derive a first circle fitted to said selected boundary points;

selecting second and third circles concentric with said first circle and respectively having diameters larger and smaller by a predetermined amount than the diameter of said first circle;

detecting which of the possible cardiac boundary points are located in a region between said second and third circles; and

selecting those possible cardiac boundary points detected as being located in the region between said second and third circles as cardiac boundary points.

9. The method according to claim 8, comprising:

measuring the size of the cardiac contour fitted to said cardiac boundary points; and

determining if the size measured in the preceding step is greater than a predetermined value, and if so, reducing by a predetermined correction factor the vertical distance from the centroid of the area bounded by the cardiac contour of at least part of that portion of the cardiac contour beneath the centroid.

10. The method according to claim 9, comprising:

predicting by linear extrapolation from the two highest boundary points on a selected side of the cardiac contour where a next predicted cardiac boundary point would be;

determining whether the first rejected possible cardiac boundary point outside said second circle and adjacent the selected side of the cardiac contour is within a predetermined distance of said next predicted cardiac boundary point, and if so, redefining said first rejected possible cardiac boundary point as a cardiac boundary point; and

repeating said step of fitting said model function so that said model function is fitted to said detected cardiac boundary points including the previously first rejected boundary point.

11. The method according to claim 8, comprising:

predicting by linear extrapolation from the two highest boundary points on a selected side of the cardiac contour where a next predicted cardiac boundary point would be;

determining whether the first rejected possible cardiac boundary point outside said second circle and adjacent the selected side of the cardiac contour is within a predetermined distance of said next predicted cardiac boundary point, and if so, redefining said first rejected possible cardiac boundary point as a cardiac boundary point; and

repeating said step of fitting said model function so that said model function is fitted to said detected cardiac boundary points including the previously first rejected boundary point.

12. The method according to claim 11, comprising:

selecting said shift-variant sinusoidal function f₁ (θ

) defining by the following relationship

space="preserve" listing-type="equation">f.sub.1 (θ

)=r.sub.0 +r.sub.1 cos {2(θ

-φ

)-α

g(θ

-α

)},

where r_o and r₁ are coefficients corresponding to the mean radial distance from center and the maximum variation from the mean radial distance, respectively, when the cardiac contour is expressed in a polar coordinate system and θ

is the angle represented in said polar coordinate system, and α and

φ

are computed coefficients, wherein coefficients r₀, r₁, φ and

α

are determined by means of a least squares method, and where g(θ

) is a phase term which is a predetermined function of angle θ

.

13. The method according to claim 1, further comprising:

displaying the completed cardiac contour superimposed on a chest image derived from the chest radiograph.

14. The method according to claim 13, further comprising:

deriving at least one predetermined cardiac dimensional parameter from the completed cardiac contour;

deriving at least one predetermined thoracic dimensional parameter related to a dimension of the lungs from the chest radiograph; and

determining a ratio of the at least one predetermined cardiac dimensional parameter relative to the at least one predetermined thoracic dimensional parameter.

15. The method according to claim 14, comprising:

displaying at least one of said at least one predetermined cardiac dimensional parameter, said at least one predetermined thoracic dimensional parameter, and said ratio of said at least one predetermined cardiac dimensional parameter relative to said at least one predetermined thoracic dimensional parameter.

16. The method according to claim 1, further comprising:

deriving at least one predetermined cardiac dimensional parameter from the completed cardiac contour;

deriving at least one predetermined thoracic dimensional parameter related to a dimension of the lungs from the chest radiograph; and

determining a ratio of the at least one predetermined cardiac dimensional parameter relative to the at least one predetermined thoracic dimensional parameter.

17. A method for determining a cardiac contour of a subject from a chest radiograph of the subject, comprising:

detecting plural right and left cardiac boundary points in the cardiac contour;

fitting a predetermined model function to the detected cardiac boundary points to derive a completed cardiac contour based on the fitted model function;

using as said model function a function expressed by the summation of a limited number of trigonometrical functions equivalent to a partial summation of a Fourier series; and

producing a representation of the completed cardiac contour.

18. The method according to claim 17, comprising:

selecting as said model function a function f₂ (θ

) given by, f₂ (θ

)=r₀ +r₂ cos 2(θ

-φ

₂)+r₃ cos 3(θ

-φ

₃), where r_o, r₂, r₃, φ

₂ and φ

₃ are constants determined by fitting the cardiac boundary points by means of the least squares method.

19. A method for determining a cardiac contour of a subject from a chest radiograph of the subject, comprising:

detecting plural right and left cardiac boundary points in the cardiac contour;

fitting a predetermined model function to the detected cardiac boundary points to derive a completed cardiac contour based on the fitted model function; and

producing a representation of the completed cardiac contour;

wherein said step of detecting plural right and left cardiac boundary points comprises;

determining in said chest radiograph a cardiac rectangle including data from the heart, lungs and diaphragm edges;

selecting plural horizontal profiles of the data in said cardiac rectangle, determining the first derivative of selected portions of each horizontal profile, detecting a maximum value of the first derivative in those selected portions of each horizontal profile, and determining the location of each first derivative maximum value to be a possible cardiac boundary point;

selecting plural vertical profiles of the data in said cardiac rectangle, determining the first derivative of selected portions of each vertical profile, detecting a maximum value of the first derivative in those selected portions of each vertical profile, and determining the location of each first derivative maximum value obtained from the selected portions of the vertical profiles as diaphragm edge points; and

determining whether any of said possible cardiac boundary points overlap said diaphragm points, and if so, rejecting those possible cardiac boundary points which overlap said diaphragm edge points as not being cardiac boundary points.

20. The method according to claim 19, comprising:

selecting plural of those possible cardiac boundary points, which are not overlapped by said diaphragm edge points and which are adjacent said diaphragm edge points, as cardiac boundary points;

fitting said selected cardiac boundary points to a predetermined model function in the form of a circle using a least squares method to derive a first circle fitted to said selected boundary points;

selecting second third circles concentric with said first circle and respectively having diameters larger and smaller by a predetermined amount than the diameter of said first circle;

detecting which of the non-overlapped possible cardiac boundary points are located in a region between said second and third circles; and

selecting those non-overlapped possible cardiac boundary points detected as being located in the region between said second and third circles as cardiac boundary points.

21. The method according to claim 20, comprising:

measuring the size of the cardiac contour fitted to said cardiac boundary points; and

determining if the size measured in the preceding step is greater than a predetermined value, and if so, reducing by a predetermined correction factor the vertical distance from the centroid of the area bounded by the cardiac contour of at least part of that portion of the cardiac contour beneath the centroid.

22. The method according to claim 21, comprising:

predicting by linear extrapolation from the two highest boundary points on a selected side of the cardiac contour where a next predicted cardiac boundary point would be;

determining whether the first rejected possible cardiac boundary point outside said second circle and adjacent the selected side of the cardiac contour is within a predetermined distance of said next predicted cardiac boundary point, and if so, redefining said first rejected possible cardiac boundary point as a cardiac boundary point; and

repeating said step of fitting said model function so that said model function is fitted to said detected cardiac boundary points including the previously first rejected boundary point.

23. The method according to claim 22, comprising:

using a shift-variant sinusoidal function as said predetermined model function.

24. The method according to claim 23, comprising:

selecting said shift-variant sinusoidal function f₁ (θ

) defined by the following relationship

space="preserve" listing-type="equation">f.sub.1 (θ

)-r.sub.0 +r.sub.1 cos {2(θ

-φ

)-α

g(θ

-φ

)},

where r_o and r₁ are coefficients corresponding to the mean radial distance from center and the maximum variation from the mean radial distance, respectively, when the cardiac contour is expressed in a polar coordinate system and θ

is the angle represented in said polar coordinate system, and α and

φ

are computed coefficients, wherein coefficients r₀, r₁, φ and

α

are determined by means of a least squares method, and where g(θ

) is a phase term which is a predetermined function of angle θ

.

25. The method according to claim 20, comprising:

predicting by linear extrapolation from the two highest boundary points on a selected side of the cardiac contour where a next predicted cardiac boundary point would be;

determining whether the first rejected possible cardiac boundary point outside said second circle and adjacent the selected side of the cardiac contour is within a predetermined distance of said next predicted cardiac boundary point, and if so, redefining said first rejected possible cardiac boundary point as a cardiac boundary point; and

repeating said step of fitting said model function so that said model function is fitted to said detected cardiac boundary points including the previously first rejected boundary point.

26. A system for determining a cardiac contour of a subject from a chest radiograph of the subject, comprising:

means for producing digital data corresponding to the chest radiograph;

means for processing the digital data to detect plural right and left cardiac boundary points in the cardiac contour;

means for fitting a predetermined model function to the detected cardiac boundary points to derive a completed cardiac contour based on the fitted model function; and

means for producing a representation of the completed cardiac contour;

wherein said fitting means fits a shift-variant sinusoidal function to the detected cardiac boundary points.

27. The system according to claim 26, wherein said shift-variant sinusoidal function f₁ (θ

) is defined by the following relationship

space="preserve" listing-type="equation">f.sub.1 (θ

)=r.sub.0 +r.sub.1 cos {2(θ

-φ

)-α

g(θ

-φ

)},

where r_o and r₁ are coefficients corresponding to the mean radial distance from center and the maximum variation from the mean radial distance, respectively, when the cardiac contour is expressed in a polar coordinate system and a θ

is the angle represented in said polar coordinate system, and α and

φ

are computed coefficients, wherein coefficients r₀, r₁, φ and

α

are determined by means of a least squares method, and where g(θ

) is a phase term which is a predetermined function of angle θ

.

28. The system according to claim 27, wherein:

the phase term g(θ

) of said shift-variant sinusoidal function is selected from the following group consisting of;

space="preserve" listing-type="equation">g.sub.1 (θ

)=cos θ

,

space="preserve" listing-type="equation">g.sub.2 (θ

)=sin (θ

),

space="preserve" listing-type="equation">g.sub.3 (θ

)=1-|θ

|/π

,

space="preserve" listing-type="equation">g.sub.4 (θ

)=|θ

|/π

,

space="preserve" listing-type="equation">g.sub.5 (θ

)=1-(θ

/π

).sup.2,

and

space="preserve" listing-type="equation">g.sub.6 (θ

)=(θ

/π

).sup.2.

29. The system according to claim 26, wherein said step of detecting plural right and left cardiac boundary points comprises:

means for obtaining edge gradients (G_hx, G_hy) in two orthogonal directions from pixel values located in plural narrow horizontal bands of pixels in said chest radiograph; and

means for selecting cardiac boundary points based on the obtained edge gradients.

30. The system according to claim 29, wherein said means for selecting cardiac boundary points comprises:

means for determining a cardiac midline (MDL) dividing the cardiac region of the chest radiograph into right and left portions;

means for determining an overall edge gradient G_h and an edge gradient orientation angle θ

_h based on the obtained edge gradients (G_hx, G_hy), where ##EQU5## for at least selected portions of each narrow horizontal band of pixels;

means for selecting as possible cardiac boundary points on opposite sides of the cardiac midline MDL the point in each narrow horizontal band of pixels at which G_h is a maximum;

means for selecting as a lowest cardiac boundary point from the possible cardiac boundary points that point of the possible cardiac boundary points which exhibits a minimum peak in orientation angle θ

_h ; and

means for rejecting as not being cardiac boundary points those possible cardiac boundary points located below said lowest boundary point.

31. The system according to claim 30, comprising:

means for detecting diaphragm edge points, including;

means for obtaining edge gradients (G_vx, G_vy) in two orthogonal directions from pixel values located in plural narrow vertical bands of pixels in said chest radiograph; and

means for selecting diaphragm edge points based on the edge gradients obtained in the preceding step.

32. The system according to claim 31, wherein said means for selecting diaphragm edge points comprises:

means for determining an overall edge gradient (G_v) based on the edge gradients (G_vx, G_vy), where ##EQU6## for at least selected portions of each narrow vertical band of pixels;

means for selecting as a prospective diaphragm edge point in each narrow vertical band of pixels that point at which G_v is maximum; and

means for connecting said prospective diaphragm edge points with a third order polynomial curve which defines a diaphragm edge.

33. The system according to claim 32, comprising:

means for selecting plural of those possible cardiac boundary points, which are adjacent said diaphragm edge points as cardiac boundary points;

means for fitting said selected cardiac boundary points to a predetermined model function in the form of a circle using a least squares method to derive a first circle fitted to said selected boundary points;

means for selecting second and third circles concentric with said first circle and respectively having diameters larger and smaller by a predetermined amount than the diameter of said first circle;

means for detecting which of the possible cardiac boundary points are located in a region between said second and third circles; and

means for selecting those possible cardiac boundary points detected as being located in the region between said second and third circles as cardiac boundary points.

34. The system according to claim 33, comprising:

means for measuring the size of the cardiac contour fitted to said cardiac boundary points; and

means for determining if the size measured in the preceding step is greater than a predetermined value, and if so, reducing by a predetermined correction factor the vertical distance from the centroid of the area bounded by the cardiac contour of at least part of that portion of the cardiac contour beneath the centroid.

35. The system according to claim 34, comprising:

means for predicting by linear extrapolation from the two highest boundary points on a selected side of the cardiac contour where a next predicted cardiac boundary point would be;

means for determining whether the first rejected possible cardiac boundary point outside said second circle and adjacent the selected side of the cardiac contour is within a predetermined distance of said next predicted cardiac boundary point, and if so, redefining said first rejected possible cardiac boundary point as a cardiac boundary point; and

means for repeating fitting of said model function so that said model function is fitted to said detected cardiac boundary points including the previously first rejected boundary point.

36. The system according to claim 33, comprising:

means for predicting by linear extrapolation from the two highest boundary points on a selected side of the cardiac contour where a next predicted cardiac boundary point would be;

means for determining whether the first rejected possible cardiac boundary point outside said second circle and adjacent the selected side of the cardiac contour is within a predetermined distance of said next predicted cardiac boundary point, and if so, redefining said first rejected possible cardiac boundary point as a cardiac boundary point; and

means for repeating fitting of said model function so that said model function is fitted to said detected cardiac boundary points including the previously first rejected boundary point.

37. The system according to claim 26, further comprising:

means for displaying the completed cardiac contour superimposed on a chest image derived from the chest radiograph.

38. The system according to claim 37, further comprising:

means for deriving at least one predetermined cardiac dimensional parameter from the completed cardiac contour;

means for deriving at least one predetermined thoracic dimensional parameter related to a dimension of the lungs from the chest radiograph; and

means for determining a ratio of the at least one predetermined cardiac dimensional parameter relative to the at least one predetermined thoracic dimensional parameter.

39. The system according to claim 38, comprising:

means for displaying at least one of said at least one predetermined cardiac dimensional parameter, said at least one predetermined thoracic dimensional parameter, and said ratio of said at least one predetermined cardiac dimensional parameter relative to said at least one predetermined thoracic dimensional parameter.

40. The system according to claim 26, further comprising:

means for deriving at least one predetermined cardiac dimensional parameter from the completed cardiac contour;

means for deriving at least one predetermined thoracic dimensional parameter related to a dimension of the lungs from the digital data; and

means for determining a ratio of the at least one predetermined cardiac dimensional parameter relative to the at least one predetermined thoracic dimensional parameter.

41. A system for determining a cardiac contour of a subject from a chest radiograph of the subject, comprising:

means for producing digital data corresponding to the chest radiograph;

means for processing the digital data to detect plural right and left cardiac boundary points in the cardiac contour;

means for fitting a predetermined model function to the detected cardiac boundary points to derive a completed cardiac contour based on the fitted model function; and

means for producing a representation of the completed cardiac contour;

wherein said fitting means fits a function expressed by the summation of a limited number of trigonometrical functions equivalent to a partial summation of a Fourier series to said detected cardiac boundary points.

42. The system according to claim 41, wherein said model function is a function f₂ (θ

) given by, f₂ (θ

)=r₀ +r₂ cos 2(θ

-φ

₂)+r₃ cos 3(θ

-φ

₃), where r_o, r₂, r₃, φ

₂ and φ

₃ are constants determined by fitting the cardiac boundary points by means of the least squares method.

43. A system for determining a cardiac contour of a subject from a chest radiograph of the subject, comprising:

means for producing digital data corresponding to the chest radiograph;

means for processing the digital data to detect plural right and left cardiac boundary points in the cardiac contour; and

means for fitting a predetermined model function to the detected cardiac boundary points to derive a completed cardiac contour based on the fitted model function; and

means for producing a representation of the completed cardiac contour;

wherein said means for detecting plural right and left cardiac boundary points comprises;

means for determining from said digital data a cardiac rectangle including data from the heart, lungs and diaphragm edges;

means for selecting plural horizontal profiles of the data in said cardiac rectangle, determining the first derivative of selected portions of each horizontal profile, detecting a maximum value of the first derivative in those selected portions of each horizontal profile, and determining the location of each first derivative maximum value to be a possible cardiac boundary point;

means for selecting plural vertical profiles of the data in said cardiac rectangle, determining the first derivative of selected portions of each vertical profile, detecting a maximum value of the first derivative in those selected portions of each vertical profile, and determining the location of each first derivative maximum value obtained from the selected portions of the vertical profiles as diaphragm edge points; and

means for determining whether any of said possible cardiac boundary points overlap said diaphragm points, and if so, rejecting those possible cardiac boundary points which overlap said diaphragm edge points as not being cardiac boundary points.

44. The system according to claim 43, comprising:

means for selecting plural of those possible cardiac boundary points, which are not overlapped by said diaphragm edge points and which are adjacent said diaphragm edge points, as cardiac boundary points;

means for fitting said selected cardiac boundary points to a predetermined model function in the form of a circle using a least squares method to derive a first circle fitted to said selected boundary points;

means for selecting second and third circles concentric with said first circle and respectively having diameters larger and smaller by a predetermined amount than the diameter of said first circle;

means for detecting which of the non-overlapped possible cardiac boundary points are located in a region between said second and third circles; and

means for selecting those non-overlapped possible cardiac boundary points detected as being located in the region between said second and third circles as cardiac boundary points.

45. The system according to claim 44, comprising:

means for measuring the size of the cardiac contour fitted to said cardiac boundary points; and

means for determining if the measured cardiac size is greater than a predetermined value, and if so, reducing by a predetermined correction factor the vertical distance from the centroid of the area bounded by the cardiac contour of at least part of that portion of the cardiac contour beneath the centroid.

46. The system according to claim 45, comprising:

means for predicting by linear extrapolation from the two highest boundary points on a selected side of the cardiac contour where a next predicted cardiac boundary point would be;

means for determining whether the first rejected possible cardiac boundary point outside said second circle and adjacent the selected side of the cardiac contour is within a predetermined distance of said next predicted cardiac boundary point, and if so, redefining said first rejected possible cardiac boundary point as a cardiac boundary point; and

means for repeating fitting of said model function so that said model function is fitted to said detected cardiac boundary points including the previously first rejected boundary point.

47. The system according to claim 46, wherein a shift-variant sinusoidal function is used as said predetermined model function.

48. The system according to claim 44, comprising:

means for predicting by linear extrapolation from the two highest boundary points on a selected side of the cardiac contour where a next predicted cardiac boundary point would be;

means for determining whether the first rejected possible cardiac boundary point outside said second circle and adjacent the selected side of the cardiac contour is within a predetermined distance of said next predicted cardiac boundary point, and if so, redefining said first rejected possible cardiac boundary point as a cardiac boundary point; and

means for repeating said fitting of said model function so that said model function is fitted to said detected cardiac boundary points including the previously first rejected boundary point.

49. A system for determining a cardiac contour of a subject from a chest radiograph of the subject, comprising:

means for displaying an image derived from the chest radiograph;

means for selecting from the displayed image plural right and left cardiac boundary points in the cardiac contour;

means for fitting a predetermined model function to the selected cardiac boundary points to derive a completed cardiac contour based on the fitted model function; and

means for producing a representation of the complete cardiac contour;

wherein said fitting means fits to the detected cardiac boundary points a predetermined model function selected from the group consisting of a shift-variant sinusoidal function and a function expressed by a summation of a limited number of trigonometrical functions equivalent to a partial summation of a Fourier series.

50. A method for determining a cardiac contour of a subject from a chest radiograph of the subject, comprising:

defining in said radiograph a polar coordinate system having an origin at a predetermined point in said radiograph;

detecting in said radiograph polar coordinate system polar coordinates of plural right and left cardiac boundary points in the cardiac contour;

fitting a predetermined model function to the detected polar coordinates of the plural right and left cardiac boundary points;

using as said predetermined model function a function selected from the group consisting of a shift-variant sinusoidal function and a function expressed by a summation of a limited number of trigonometrical functions equivalent to a partial summation of a Fourier series; and

producing a representation of the completed cardiac contour based on said model function fitted in said fitting step.

51. A system for determining a cardiac contour of a subject from a chest radiograph of the subject, comprising:

means for defining in said radiograph a polar coordinate system having an origin at a predetermined point in said radiograph and for detecting in said radiograph polar coordinate system polar coordinates of plural right and left cardiac boundary points in the cardiac contour;

means for fitting a predetermined model function to the detected polar coordinates of the plural right and left cardiac boundary points; and

means for producing a representation of the completed cardiac contour based upon said model function fitted by said fitting means;

wherein said fitting means fits to the detected cardiac boundary points a predetermined model function selected from the group consisting of a shift-variant sinusoidal function and a function expressed by a summation of a limited number of trigonometrical functions equivalent to a partial summation of a Fourier series.