Method for segmentation of IVUS image sequences

• US 20050249391A1
• Filed: 05/10/2004
• Published: 11/10/2005
• Est. Priority Date: 05/10/2004
• Status: Active Grant

• Current Assignee: MediGuide Ltd (Abbott Laboratories)
• Original Assignee: MediGuide Ltd (Abbott Laboratories)
• Inventors: Sidlesky, Avishav, Sudarsky, Oded, Kimmel, Ron, Petrank, Yael
• Granted Patent: US 7,397,935 B2
• Time in Patent Office: Days
• US Class Current: 382/128
• CPC Class Codes: G06T 2207/10132 Ultrasound image
• CPC Class Codes: G06T 2207/30101 Blood vessel; Artery; Vein; Vascular
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/143 involving probabilistic approaches, e.g. Markov random field [MRF] modelling

Abstract

Method for performing segmentation of an interior vessel within the body of a patient, the method including obtaining a sequence of intravascular ultrasound images of an interior vessel and dividing the sequence into batches, detecting uninformative regions in each of the batches, the uninformative regions arising from an acoustic shadow cast by guide wire and by calcified plaque within the interior vessel, extracting a preliminary outer boundary of the interior vessel, tracking images in each of the batches to counter various distortions, performing statistical analysis and spatial integration on each of the batches to obtain a classification of blood and tissue regions, extracting a secondary outer boundary of the interior vessel utilizing the classification of blood and tissue regions and refining result, and extracting the inner boundary of the interior vessel based on the classification of blood and tissue regions.

Claims

1. A method for performing segmentation of an interior vessel within the body of a patient to detect the inner boundary of said vessel, the method comprising the procedures of:

obtaining a sequence of intravascular ultrasound (IVUS) images of an interior vessel;

tracking images in said sequence to counter various distortions;

performing statistical analysis and spatial integration on said sequence to obtain a classification of blood and tissue regions; and

extracting inner boundary of said interior vessel based on said classification of blood and tissue regions.

2. The method according to claim 1, further comprising the procedures of:

scaling down said sequence of IVUS images, at least before said procedure of performing statistical analysis and spatial integration; and

scaling up said extracted inner boundary.

7. The method according to claims 1, or 6, wherein said sequence of IVUS images is divided into batches.

8. The method according to claims 1, 3, or 6, wherein said interior vessel is a coronary artery.

10. The method according to claims 1, 3, or 6, further comprising the procedure of:

performing roll stabilization on said sequence of IVUS images.

11. The method according to claims 1, 3, or 6, further comprising the procedure of:

detecting uninformative regions in said sequence of IVUS images, said uninformative regions arising from an acoustic shadow cast by guide wire and by calcified plaque within said interior vessel.

12. The method according to claim 11, wherein said procedure of detecting uninformative regions in said sequence of IVUS images, comprises the procedures of:

detecting the location of said guide wire in said sequence of IVUS images;

identifying regions of calcified plaque in said sequence of IVUS images; and

integrating said guide wire data with said regions of calcified plaque data.

13. The method according to claim 12, wherein said procedure of detecting the location of said guide wire in said sequence of IVUS images, comprises the procedures of:

locating the radial ray with a maximum difference between the brightness at the bottom of said radial ray and the standard deviation in the rest of said radial ray, by searching all radial rays of every image in said sequence of IVUS images, resulting in a first sequence of radial rays;

locating the radial ray with a maximum difference between the brightness at the bottom of said radial ray and the standard deviation in the rest of said radial ray, by searching only radial rays of every image in said sequence of IVUS images which fit the profile of a typical guide wire radial ray, resulting in a second sequence of radial rays;

locating the radial distance with a maximum intensity within said found radial ray, for said first sequence and for said second sequence;

locating the radial ray with a nearest local maximum intensity within said found radial distance, for said first sequence;

excluding anomalous images from said first sequence and from said second sequence;

applying a median filter to said first sequence and to said second sequence;

fitting a spline through guide wire center radial rays and excluding outlying images from said first sequence and from said second sequence, in two rounds;

fitting a spline through guide wire center radial rays of remaining images in said first sequence and in said second sequence;

calculating dispersion in said first sequence and in said second sequence, said dispersion being the width of said acoustic shadow cast by said guide wire; and

selecting a sequence from one of said first sequence and said second sequence by lowest dispersion value, only if said dispersion value is below an allowed threshold.

14. The method according to claim 12, wherein said procedure of detecting the location of said guide wire in said sequence of IVUS images, comprises the procedures of:

locating the column with a maximum difference between the brightness at the bottom of said column and the standard deviation in the rest of said column, by searching all columns of every image in said sequence of IVUS images, resulting in a first sequence of columns;

locating the column with a maximum difference between the brightness at the bottom of said column and the standard deviation in the rest of said column, by searching only columns of every image in said sequence of IVUS images which fit the profile of a typical guide wire column, resulting in a second sequence of columns;

locating the row with a maximum intensity within said found column, for said first sequence and for said second sequence;

locating the column with a nearest local maximum intensity within said found row, for said first sequence;

excluding anomalous images from said first sequence and from said second sequence;

applying a median filter to said first sequence and to said second sequence;

fitting a spline through guide wire center columns and excluding outlying images from said first sequence and from said second sequence, in two rounds;

fitting a spline through guide wire center columns of remaining images in said first sequence and in said second sequence;

calculating dispersion in said first sequence and in said second sequence, said dispersion being the width of said acoustic shadow cast by said guide wire; and

selecting a sequence from one of said first sequence and said second sequence by lowest dispersion value, only if said dispersion value is below an allowed threshold

wherein said IVUS images are provided in polar coordinate form, and wherein the horizontal axis of said IVUS images corresponds to the angular dimension.

15. The method according to claim 12, wherein said procedure of identifying regions of calcified plaque in said sequence of IVUS images, comprises the procedures of:

applying a morphological closing operation to every image in said sequence of IVUS images,

assigning tentative calcium pixels based on pixel intensity characteristics, resulting in a binary image of true and false pixels;

applying a morphological dilation operation to said resulting binary image;

determining boundary of calcified plaque regions; and

ensuring temporal continuity of said calcified plaque regions.

16. The method according to claim 7, further comprising the procedures of:

detecting stents in each of said batches; and

determining borders of said stents in said sequence of IVUS images.

17. The method according to claim 16, wherein said procedure of detecting stents in each of said batches, comprises the procedures of:

applying a morphological closing operation to every image in each of said batches;

partitioning each image into sub-regions and locating pixels with outlying intensities in each of said sub-regions;

assigning tentative stent pixels based on pixel intensity characteristics, resulting in a binary image of true and false pixels;

applying a morphological dilation operation to said resulting binary image;

eliminating falsely identified stent pixels from said binary image;

determining correlation value between consecutive images in each of said batches; and

determining if a stent exists in a batch for each of said batches.

18. The method according to claims 1, 3, or 6, further comprising the procedure of:

extracting a preliminary outer boundary of said interior vessel.

19. The method according to claim 18, wherein said extracted preliminary outer boundary is detected using a dynamic programming algorithm.

20. The method according to claims 1 or 6, wherein said procedure of tracking images involves finding the relative movement between adjacent image frames in said sequence of IVUS images by calculating motion vectors, each of said motion vectors being the offset between adjacent image frames over a sparse grid of image rectangles.

21. The method according to claims 1 or 6, wherein said procedure of tracking images is performed using a two-dimensional Kanade-Lucas algorithm.

22. The method according to claim 7, further comprising the procedures of:

estimating roll offsets between each pair of consecutive images in each of said batches; and

accumulating said roll offsets in said sequence of IVUS images.

23. The method according to claim 22, wherein said procedure of estimating roll offsets is performed based on statistics of the angular component of the motion vectors obtained in said procedure of tracking images in each of said batches to counter various distortions, said motion vectors being the offset between adjacent image frames over a sparse grid of image rectangles.

24. The method according to claims 1 or 6, wherein said classification of blood and tissue regions is based on the presence or absence of blood speckle in said images.

25. The method according to claims 1 or 6, wherein said procedure of performing statistical analysis and spatial integration involves calculating statistical properties relating to the intensities of a pixel over a series of said images.

26. The method according to claim 25, wherein said statistical properties includes properties selected from the list consisting of:

average intensity;

total variation;

number of extrema;

variance;

standard deviation;

skewness; and

kurtosis.

27. The method according to claim 25, wherein said statistical properties are calculated over a series of pixel intensity values taken from corresponding pixel coordinates in consecutive IVUS images.

28. The method according to claim 25, wherein said statistical properties are calculated over a series of pixel intensity values taken from adjacent pixel coordinates in an individual IVUS image.

29. The method according to claim 25, wherein said procedure of tracking images involves finding the relative movement between adjacent image frames in said sequence of IVUS images by calculating motion vectors, each of said motion vectors being the offset between adjacent image frames over a sparse grid of image rectangles, wherein said statistical properties are calculated for pixel coordinates that are shifted between said images in accordance with said motion vectors.

30. The method according to claim 25, wherein said statistical properties are learned from manually traced images, said manually traced images being IVUS images where the inner and outer boundaries of said interior vessel have been manually traced by a user.

31. The method according to claim 30, wherein two histograms are constructed based on said manually traced images, one of said histograms indicating the number of pixels in said manually traced images that have a given value of said statistical properties in the inner layer of said interior vessel, the other of said histograms indicating the number of pixels in said manually traced images that have a given value of said statistical properties in the outer layer of said interior vessel.

32. The method according to claim 31, wherein a classification table is generated from said histograms, said classification table being constructed according to the difference between said histograms.

33. The method according to claim 32, wherein the radius domain of said sequence of IVUS images is divided into several distinct bands, and a different classification table is generated for each of said bands.

34. The method according to claim 32, wherein pixels are classified as blood in accordance with said classification table, resulting in a sparse binary image of blood pixels.

35. The method according to claim 34, wherein spatial integration is performed on said sparse binary image of blood pixels resulting in a blood mask, said blood mask being a binary image with contiguous blood pixels.

36. The method according to claim 35, wherein said spatial integration involves performing a morphological closing operation, geometric consideration, and smoothing, on said sparse binary image of blood pixels.

37. The method according to claims 1 or 6, wherein said procedure of performing statistical analysis and spatial integration, is performed on temporally sub-sampled images in said sequence of IVUS images.

38. The method according to claim 37, wherein said temporally sub-sampled images are non-uniformly distributed over time.

43. The method according to claim 7, further comprising the procedures of:

identifying bifurcation sectors in each of said batches; and

computing bifurcation regions in said sequence of IVUS images.

44. The method according to claim 43, wherein said procedure of identifying bifurcation sectors in each of said batches is based on blood speckles detected linearly along a curve at a constant offset beyond said extracted outer boundary.

45. The method according to claim 43, wherein said procedure of identifying bifurcation sectors in each of said batches is performed excluding said detected uninformative regions arising from an acoustic shadow cast by guide wire and by calcified plaque within said interior vessel.

46. The method according to claim 43, wherein said procedure of computing said bifurcation regions in said sequence of IVUS images is assisted by bifurcation regions identified via at least one medical imaging modality.

47. The method according to claim 46, wherein said at least one medical imaging modality includes planar or bi-planar fluoroscopic angiography.

48. The method according to claims 1, 3, or 6, further comprising the procedure of:

manually editing output of said segmentation in selected images by a user.

49. The method according to claim 48, wherein said procedure of manually editing output of said segmentation involves integrating manual editing on the detected boundaries in a given image by said user, while adjusting the boundaries in adjacent images to seamlessly integrate said manual editing.

50. The method according to claim 49, wherein said procedure of manually editing output of said segmentation further considers the relative movement of each pixel in a manually edited boundary relative to the original boundary.

51. The method according to claim 48, wherein said procedure of manually editing output of said segmentation serves as feedback for an additional execution of said segmentation method, by forcing the output at desired locations to be identical to anchor points specified by said user in said manual edited images.

3. A method for performing segmentation of an interior vessel within the body of a patient to detect the outer boundary of said vessel, the method comprising the procedures of:

obtaining a sequence of intravascular ultrasound (IVUS) images of an interior vessel;

extracting an outer boundary of said interior vessel utilizing dynamic programming,

wherein images in said sequence of IVUS images are spatially associated with one another.

4. The method according to claim 3, further comprising the procedures of:

scaling down said sequence of IVUS images, at least before said procedure of extracting an outer boundary; and

scaling up said extracted outer boundary.

5. The method according to claim 3, wherein said sequence of IVUS images is divided into batches.

39. The method according to claims 3 or 6, wherein said procedure of extracting an outer boundary of said interior vessel, is performed based on said blood mask.

40. The method according to claims 3 or 6, wherein said procedure of extracting an outer boundary of said interior vessel involves approximating said extracted outer boundary in said images with the nearest matching ellipse.

41. The method according to claims 3 or 6, wherein said dynamic programming algorithm is performed using a Viterbi algorithm.

42. The method according to claims 3 or 6, wherein said dynamic programming algorithm is executed in an additional pass in a reverse direction.

6. A method for performing segmentation of an interior vessel within the body of a patient to detect both the inner boundary and the outer boundary of said vessel, the method comprising the procedures of:

obtaining a sequence of intravascular ultrasound (IVUS) images of an interior vessel;

tracking images in said sequence of IVUS images to counter various distortions;

performing statistical analysis and spatial integration on said sequence of IVUS images to obtain a classification of blood and tissue regions;

extracting inner boundary of said interior vessel based on said classification of blood and tissue regions; and

extracting an outer boundary of said interior vessel utilizing dynamic programming.

9. The method according to claim 6, further comprising the procedures of:

scaling down said sequence of IVUS images, at least before said procedure of performing statistical analysis and spatial integration; and

scaling up said extracted outer boundary and said extracted inner boundary.

52. The method according to claim 6, further comprising the procedure of:

displaying output of said segmentation with respect to level of confidence measures.

53. The method according to claim 52, wherein said level of confidence measures is represented as a confidence interval about the computed value of said output of said segmentation, such that the actual value of said output of said segmentation lies within said interval with a predetermined probability.

54. The method according to claim 52, wherein said level of confidence measures is represented as the probability that the actual value of said output of said segmentation lies within some predetermined interval about the computed value of said output of said segmentation.

55. A method for identifying regions of calcified plaque in an interior vessel within the body of a patient, the method comprising the procedures of:

obtaining a sequence of intravascular ultrasound (IVUS) images of an interior vessel;

applying a morphological closing operation to every image in said sequence of IVUS images;

assigning tentative calcium pixels based on pixel intensity characteristics, resulting in a binary image of true and false pixels;

applying a morphological dilation operation to said resulting binary image; and

determining boundary of calcified plaque regions.

56. The method according to claims 55, further including the procedure of:

ensuring temporal continuity of said calcified plaque regions.

58. The method according to claims 55 or 57, wherein said interior vessel is a coronary artery.

57. A method for detecting stents in an interior vessel within the body of a patient, the method comprising the procedures of:

obtaining a sequence of intravascular ultrasound (IVUS) images of an interior vessel and dividing said sequence into batches;

applying a morphological closing operation to every image in each of said batches;

partitioning each image into sub-regions and locating pixels with outlying intensities in each of said sub-regions;

assigning tentative stent pixels based on pixel intensity characteristics, resulting in a binary image of true and false pixels;

applying a morphological dilation operation to said resulting binary image;

eliminating falsely identified stent pixels from said binary image;

determining correlation value between consecutive images in each of said batches; and

determining if a stent exists in a batch for each of said batches.