Computer-assisted diagnosis method and system for automatically determining diagnostic saliency of digital images
First Claim
1. A computer assisted diagnosis system for automatically determining diagnostic saliency of regions in a digital image comprised of a plurality of pixels, the system comprising:
- a memory unit;
a plurality of filters stored in said memory unit, each of the plurality of filters designed to identify a specific type of diagnostic finding;
a virtual window for defining regions in the image at which each of the plurality of filters is applied;
a set of training image patches corresponding to typical appearances of the specific type of diagnostic finding;
a distance measure set corresponding to the training image patches and the regions in the image defined by the virtual window;
a feature set corresponding to the distance measure set; and
a processor for applying each of the plurality of filters to the image to compute distances between the regions in the image defined by the virtual window and the training image patches bases on the distance measure set and the feature set, and ranking regions in the image corresponding to a particular type of diagnostic finding based on the computed distances.
4 Assignments
0 Petitions
Accused Products
Abstract
A computer-assisted diagnosis method and system are provided for automatically determining diagnostic saliency of digital images. The method includes the step of providing filters for evaluating the image. Each filter is designed to identify a specific type of diagnostic finding, and is associated with the following: a virtual window for defining regions in the image at which the filter is applied; a set of training image patches corresponding to typical appearances of the specific type of diagnostic finding; a distance measure between the training image patches and the regions in the image defined by the virtual window; and a feature set corresponding to the distance measure. The filters are applied to the image to compute distances between the regions in the image defined by the virtual window and the training image patches based on the distance measure and the feature set, for each of the plurality of filters. Regions in the image are ranked as corresponding to a particular type of diagnostic finding based the computed distances.
-
Citations
37 Claims
-
1. A computer assisted diagnosis system for automatically determining diagnostic saliency of regions in a digital image comprised of a plurality of pixels, the system comprising:
-
a memory unit;
a plurality of filters stored in said memory unit, each of the plurality of filters designed to identify a specific type of diagnostic finding;
a virtual window for defining regions in the image at which each of the plurality of filters is applied;
a set of training image patches corresponding to typical appearances of the specific type of diagnostic finding;
a distance measure set corresponding to the training image patches and the regions in the image defined by the virtual window;
a feature set corresponding to the distance measure set; and
a processor for applying each of the plurality of filters to the image to compute distances between the regions in the image defined by the virtual window and the training image patches bases on the distance measure set and the feature set, and ranking regions in the image corresponding to a particular type of diagnostic finding based on the computed distances. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21)
-
-
22. A computer assisted diagnosis method for automatically determining diagnostic saliency of locations in a digital image comprised of a plurality of pixels, the method comprising the steps of:
-
providing a plurality of filters for evaluating the image, wherein each of the plurality of filters is designed to identify a specific type of diagnostic finding;
providing a virtual window for defining regions in the image at which each of the plurality of filters is applied;
providing a set of training image patches corresponding to typical appearances of the specific type of diagnostic finding;
providing a distance measure set corresponding to the training image patches and the regions in the image defined by the virtual window;
providing a feature set corresponding to the distance measure set;
applying the plurality of filters to the image to compute distances between the regions in the image defined by the virtual window and the training image patches based on the distance measure set and the feature set, for each of the plurality of filters; and
ranking regions in the image as corresponding to a particular type of diagnostic finding based the computed distances. - View Dependent Claims (23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37)
constructing a multi-resolution image pyramid from the image;
applying some of the plurality of filters to the multi-resolution image pyramid; and
aggregating results between pyramid levels for each filter.
-
-
27. The method according to claim 22, further comprising the step of normalizing outputs of each of the plurality of filters.
-
28. The method according to claim 22, further comprising the step of adjusting sensitivities of the plurality of filters to be one of uniform and independent.
-
29. The method according to claim 22, further comprising the steps of identifying and ranking foci of attention belonging to a particular type of diagnostic finding using individual filter responses.
-
30. The method according to claim 22, further comprising the step of combining outputs of the plurality of filters into a single diagnostic saliency map for the image.
-
31. The method according to claim 22, further comprising the steps of identifying and ranking diagnostic foci of attention in the image at a given level of sensitivity, using the diagnostic saliency map.
-
32. The method according to claim 22, further comprising the step of combining the outputs of the plurality of filters using winner-take-all operators between the regions in the image defined by the virtual windows, and one of aggregation operators and winner-take-all operators between different filters.
-
33. The method according to claim 22, further comprising the step of removing salt-and-pepper type of noise from outputs of the filters.
-
34. The method according to claim 22, further comprising the step of generating a numeric diagnostic saliency score for the regions in the image defined by the virtual windows associated with the plurality of filters based on the computed distances.
-
35. The method according to claim 22, further comprising the step of combining outputs of the plurality of filters into a single diagnostic saliency map for the image, the map comprising a numeric diagnostic saliency score for each of the regions in the image defined by the virtual windows associated with the plurality of filters, the specific type of diagnostic finding at each of the regions and characteristics of the diagnostic finding.
-
36. The method according to claim 22, further comprising the step of identifying an individual training image patch based on an output of the individual training image patch.
-
37. The method according to claim 22, further comprising the step of integrating domain knowledge rules into one of normalizing outputs of the plurality of filters and generating a diagnostic saliency map for the image.
Specification