Ribcage segmentation
View Patent ↗A technique for segmenting a ribcage, e.g., in a chest radiograph, may involve determining whether or not the top of the image cuts through the lungs. The process may then proceed to segment the top and the sides of the ribcage. A dynamic programming technique may be used to perform the segmentation.
1. A method of performing ribcage segmentation in a chest image, the method comprising:
classifying a top of the chest image as having lungs adjacent to the top or not adjacent to the top;
segmenting the top of the ribcage, based on the result of said classifying, to obtain a top mask; and
segmenting at least one outer side of the ribcage, after the top of the ribcage is segmented, by using an endpoint of the top mask as a starting point,
wherein at least one of said segmenting the top of the ribcage or said segmenting at least one outer side of the ribcage comprises:
obtaining a local cost image based on the chest image; and
applying a dynamic programming algorithm to the local cost image.
2. The method of claim 1 , wherein said classifying the top of the chest image comprises:
forming a shape-index image and using the shape-index image to classify the top of the chest image.
3. The method of claim 1 , wherein said obtaining a local cost image comprises:
combining two or more cost images computed from the chest image to obtain the local cost image.
4. The method of claim 3 , wherein said two or more cost images are selected from the group consisting of: a second derivative image; a gradient convergence image; and a shape-index image.
5. The method of claim 1 , wherein said applying a dynamic programming algorithm to the local cost image comprises:
tracing a path based on a cumulative cost based on the local cost image and transition costs.
6. The method of claim 1 , further comprising:
pre-processing the chest image prior to classifying the top of the chest image, wherein said pre-processing includes at least one pre-processing technique selected from the group consisting of: adjusting image resolution; locating an approximate location of a spinal column in the chest image; and removing extraneous border content from the chest image.
7. The method of claim 1 , wherein said segmenting the top of the ribcage comprises transforming a top portion of the chest image to polar coordinates.
8. The method of claim 7 , wherein said transforming includes:
folding the top portion of the chest image across a horizontal line to obtain a folded image having vertical symmetry; and
performing a polar transformation on the folded image.
9. A computer-readable non-transitory storage medium containing instructions that, when executed by a processor, cause the processor to implement a method of performing ribcage segmentation in a chest image, the method comprising:
classifying a top of the chest image as having lungs adjacent to the top or not adjacent to the top;
segmenting the top of the ribcage, based on the result of said classifying, to obtain a top mask; and
segmenting at least one outer side of the ribcage, after the top of the ribcage is segmented, by using an endpoint of the top mask as a starting point,
wherein at least one of said segmenting the top of the ribcage or said segmenting at least one outer side of the ribcage comprises:
obtaining a local cost image based on the chest image; and
applying a dynamic programming algorithm to the local cost image.
10. The computer-readable non-transitory storage medium of claim 9 , wherein said classifying the top of the chest image comprises:
forming a shape-index image and using the shape-index image to classify the top of the chest image.
11. The computer-readable non-transitory storage medium of claim 9 , wherein said obtaining a local cost image comprises:
combining two or more cost images computed from the chest image to obtain the local cost image.
12. The computer-readable non-transitory storage medium of claim 11 , wherein said two or more cost images are selected from the group consisting of: a second derivative image; a gradient convergence image; and a shape-index image.
13. The computer-readable non-transitory storage medium of claim 9 , wherein said applying a dynamic programming algorithm to the local cost image comprises:
tracing a path based on a cumulative cost based on the local cost image and transition costs.
14. The computer-readable non-transitory storage medium of claim 9 , wherein the method further comprises:
pre-processing the chest image prior to classifying the top of the chest image, wherein said pre-processing includes at least one pre-processing technique selected from the group consisting of: adjusting image resolution; locating an approximate location of a spinal column in the chest image; and removing extraneous border content from the chest image.
15. The computer-readable non-transitory storage medium of claim 9 , wherein said segmenting the top of the ribcage comprises transforming a top portion of the chest image to polar coordinates.
16. The computer-readable non-transitory storage medium of claim 15 , wherein said transforming includes:
folding the top portion of the chest image across a horizontal line to obtain a folded image having vertical symmetry; and
performing a polar transformation on the folded image.
17. An apparatus for performing ribcage segmentation in a chest image, comprising:
means for classifying a top of the chest image as having lungs adjacent to the top or not adjacent to the top;
means for segmenting the top of the ribcage, based on the result of said classifying, to obtain a top mask; and
means for segmenting at least one outer side of the ribcage, after the top of the ribcage is segmented, by using an endpoint of the top mask as a starting point,
wherein at least one of said means for segmenting the top of the ribcage or said means for segmenting at least one outer side of the ribcage comprises:
means for obtaining a local cost image based on the chest image; and
means for applying a dynamic programming algorithm to the local cost image.
18. The apparatus of claim 17 , wherein said means for classifying the top of the chest image comprises:
means for forming a shape-index image and using the shape-index image to classify the top of the chest image.
19. The apparatus of claim 17 , wherein said means for applying a dynamic programming algorithm to the local cost image comprises:
means for tracing a path based on a cumulative cost based on the local cost image and transition costs.
20. The apparatus of claim 17 , further comprising:
means for pre-processing the chest image prior to classifying the top of the chest image, wherein said pre-processing includes at least one pre-processing technique selected from the group consisting of: adjusting image resolution; locating an approximate location of a spinal column in the chest image; and removing extraneous border content from the chest image.
21. The apparatus of claim 17 , wherein said means for segmenting the top of the ribcage is configured to transform a top portion of the chest image to polar coordinates.
22. The apparatus of claim 21 wherein said means for segmenting the top of the ribcage is configured to transform a top portion of the chest image to polar coordinates by:
folding the top portion of the chest image across a horizontal line obtain a folded image having vertical symmetry; and
performing a polar transformation on the folded image.