Image processing device, image processing method, and computer-readable recording medium for detecting abnormality from intraluminal image using integrated feature data

1. An image processing device comprising:

a processor comprising hardware, wherein the processor is configured to:

detect, from an intraluminal image obtained by imaging a lumen of a living body, an abnormality candidate region in which a tissue characteristic of the living body or an in-vivo state satisfies a predetermined condition;

calculate, from each of a plurality of regions inside the intraluminal image, a plurality of pieces of feature data including different kinds;

acquire information of the abnormality candidate region;

estimate an abnormality kind in the abnormality candidate region;

set, as a calculation parameter, one of:

number of dimensions for each kind of feature data used to calculate integrated feature data, based on the information of the abnormality candidate region; and

transform number of dimensions for the feature data, based on the information of the abnormality candidate region;

calculate the integrated feature data by integrating the plurality of pieces of feature data based on the calculation parameter; and

detect an abnormality from the intraluminal image by using the integrated feature data.

2. The image processing device according to claim 1 ,

wherein the processor is configured to select feature data to be used when calculating the integrated feature data based on the abnormality kind estimated.

3. The image processing device according to claim 2 ,

wherein the abnormality kind estimated comprises abnormality of a color in the intraluminal image, abnormality of a shape in the intraluminal image, and abnormality of texture in the intraluminal image, and

wherein the processor is configured to preferentially select, in accordance with the abnormality kind estimated, feature data corresponding to the abnormality kind estimated over feature data corresponding to other abnormality kinds.

4. The image processing device according to claim 1 ,

wherein the processor is configured to:

estimate the abnormality kind in the abnormality candidate region by a plurality of methods different from each other; and

select feature data to be used when calculating the integrated feature data by combining at least part of estimation results of the abnormality kind by the plurality of methods.

5. The image processing device according to claim 1 ,

wherein the processor is configured to:

estimate, as the abnormality kind, which one of a color, a shape and texture an abnormality exists in; and

determine the number of dimensions,

wherein the number of dimensions when an abnormality exists in the color is set smaller than the number of dimensions when an abnormality exists in the shape and/or the texture.

6. The image processing device according to claim 1 ,

wherein the processor is configured to:

estimate the abnormality kind in the abnormality candidate region based on a plurality of methods different from each other, and

perform at least any one of:

determine the number of dimensions for the integrated feature data by combining at least part of estimation results of the abnormality kind by the plurality of methods; and

determine transform number of dimensions for the feature data.

7. The image processing device according to claim 1 ,

wherein the processor is configured to calculate clarity at a boundary of the abnormality candidate region.

8. The image processing device according to claim 7 ,

wherein the processor is configured to select feature data to be used when calculating the integrated feature data.

9. The image processing device according to claim 8 ,

wherein the processor is configured to:

preferentially select at least one of color feature data and shape feature data when the boundary is clear; and

preferentially select texture feature data when the boundary is unclear.

10. The image processing device according to claim 7 ,

wherein the processor is configured to, set the number of dimensions when the boundary is clear to be smaller than the number of dimensions when the boundary is unclear.

11. The image processing device according to claim 1 ,

wherein the processor is configured to calculate information on a peripheral region of the abnormality candidate region.

12. The image processing device according to claim 11 ,

wherein the processor is configured to select feature data to be used when calculating the integrated feature data based on the information on the peripheral region of the abnormality candidate region.

13. The image processing device according to claim 12 ,

wherein the processor is configured to categorize organs in the intraluminal image into kinds.

14. The image processing device according to claim 13 ,

wherein the processor is configured to:

when the kind of the organ is the stomach or small intestine, preferentially select at least one of color feature data and texture feature data; and

when the kind of the organ is a large intestine, preferentially select shape feature data.

15. The image processing device according to claim 12 ,

wherein the processor is configured to determine whether the abnormality candidate region exists in a bubble inner region.

16. The image processing device according to claim 15 ,

wherein the processor is configured to, when the abnormality candidate region exists in a bubble inner region, preferentially select at least one of shape feature data and texture feature data.

17. The image processing device according to claim 12 ,

wherein the processor is configured to determine at least one of whether the abnormality candidate region exists in a dark region and whether the abnormality candidate region exists in a halation region.

18. The image processing device according to claim 17 ,

wherein the processor is configured to, when the abnormality candidate region exists in a dark region or a halation region, preferentially select at least one of shape feature data and texture feature data.

19. The image processing device according to claim 12 ,

wherein the processor is configured to abnormality calculate a mucosal color in the intraluminal image.

20. The image processing device according to claim 19 ,

wherein the processor is configured to:

preferentially select color feature data when the mucosal color has weak redness; and

preferentially select at least one of shape feature data and the texture feature data when the mucosal color has strong redness.

21. The image processing device according to claim 11 ,

wherein the processor is configured to categorize organs in the intraluminal image into kinds.

22. The image processing device according to claim 21 ,

wherein the processor is configured to, when determining the number of dimensions, set the number of dimensions for an organ to be examined to be larger than the number of dimensions for an organ not to be examined.

23. The image processing device according to claim 11 ,

wherein the processor is configured to determine whether the abnormality candidate region exists in a bubble inner region.

24. The image processing device according to claim 23 ,

wherein the processor is configured to, when determining the number of dimensions, set the number of dimensions when the abnormality candidate region exists in the bubble inner region to be larger than the number of dimensions when the abnormality candidate region exists a bubble outer region.

25. The image processing device according to claim 11 ,

wherein the processor is configured to determine at least one of whether the abnormality candidate region exists in a dark region and whether the abnormality candidate region exists in a halation region.

26. The image processing device according to claim 25 ,

wherein the processor is configured to, when determining the number of dimensions, set the number of dimensions when the abnormality candidate region exists in the dark region or the halation region to be larger than the number of dimensions when the abnormality candidate region exists in a non-dark region or a non-halation region.

27. The image processing device according to claim 11 ,

wherein the processor is configured to calculate a mucosal color in the intraluminal image.

28. The image processing device according to claim 27 ,

wherein the processor is configured to, when determining the number of dimensions, setting the number of dimensions when the mucosal color has strong redness to be larger than number of dimensions when the mucosal color has weak redness.

29. An image processing method comprising:

detecting, from an intraluminal image obtained by imaging a lumen of a living body, an abnormality candidate region in which a tissue characteristic of the living body or an in-vivo state satisfies a predetermined condition;

calculating, from each of a plurality of regions inside the intraluminal image, a plurality of pieces of feature data including different kinds;

acquiring information of the abnormality candidate region;

estimating an abnormality kind in the abnormality candidate region;

setting, as a calculation parameter, one of:

number of dimensions for each kind of feature data used to calculate integrated feature data, based on the information of the abnormality candidate region; and

transform number of dimensions for the feature data, based on the information of the abnormality candidate region;

calculating the integrated feature data by integrating the plurality of pieces of feature data based on the calculation parameter; and

detecting an abnormality from the intraluminal image by using the integrated feature data.

30. A non-transitory computer-readable recording medium with an executable program stored thereon, wherein the program causing a processor to execute:

detecting, from an intraluminal image obtained by imaging a lumen of a living body, an abnormality candidate region in which a tissue characteristic of the living body or an in-vivo state satisfies a predetermined condition;

calculating, from each of a plurality of regions inside the intraluminal image, a plurality of pieces of feature data including different kinds;

acquiring information of the abnormality candidate region;

estimating an abnormality kind in the abnormality candidate region;

setting, as a calculation parameter, one of:

number of dimensions for each kind of feature data used to calculate integrated feature data, based on the information of the abnormality candidate region; and

transform number of dimensions for the feature data, based on the information of the abnormality candidate region;

calculating the integrated feature data by integrating the plurality of pieces of feature data based on the calculation parameter; and

detecting an abnormality from the intraluminal image by using the integrated feature data.