Systems and methods for automated analysis of heterotopic ossification in 3D images

Patent metadata

Field	Value
Publication number	US-10813614-B2
Application number	US-201715604350-A
Country	US
Kind code	B2
Filing date	May 24, 2017
Priority date	May 24, 2017
Publication date	Oct 27, 2020
Grant date	Oct 27, 2020

Abstract

Presented herein are systems and methods that facilitate automated segmentation of 3D images of subjects to distinguish between regions of heterotopic ossification (HO) normal skeleton, and soft tissue. In certain embodiments, the methods identify discrete, differentiable regions of a 3D image of subject (e.g., a CT or microCT image) that may then be either manually or automatically classified as either HO or normal skeleton.

First claim

What is claimed is:

1. A method for automatically detecting heterotopic ossification (HO) in a 3D image of a subject, the method comprising:
(a) receiving, by a processor of a computing device, the 3D image of a subject;
(b) applying, by the processor, a global thresholding operation to the 3D image to produce an initial bone mask that identifies an initial region of interest within the image comprising a graphical representation of bone;
(c) determining, by the processor, a boundary value map using a 3D edge detection operation applied to the initial region of interest of the 3D image identified by the initial bone mask, wherein the boundary value map identifies and includes intensity values of voxels of the 3D image that correspond to boundaries where bone meets soft tissue;
(d) determining, by the processor, a bone threshold map using the initial bone mask and the boundary value map, wherein the bone threshold map comprises, for each voxel of the initial bone mask, a threshold value determined by extrapolating values of the boundary value map to voxels within the initial bone mask;
(e) determining, by the processor, a final bone mask using the bone threshold map and the 3D image;
(f) determining, by the processor, a distance map by applying a distance transform to the final bone mask;
(g) applying, by the processor, a watershed segmentation operation to the distance map to identify at least one of a set of catchment basins and watershed lines within the distance map;
(h) generating, by the processor, a first split bone mask using the final bone mask and at least one of the identified catchment basins and the watershed lines from act (g);
(i) applying, by the processor, one or more second derivative splitting filters to the 3D image to identify a set of split line voxels within the 3D image;
(j) removing, by the processor, voxels corresponding to the set of split line voxels from the first split bone mask, thereby generating a second split bone mask;
(k) determining, by the processor, a plurality of labeled split binary components of the second split bone mask via one or more morphological processing operations;
(l) performing, by the processor, a region growing operation within the final bone mask using the plurality of labeled split binary components of the second split bone mask as seeds, thereby producing a labeled final bone map; and
(m) rendering, by the processor, a graphical representation of the labeled final bone map.

2. The method of claim 1 , wherein act (b) comprises determining, by the processor, a global threshold value using intensities of voxels of the 3D image.

3. The method of claim 2 , wherein the global threshold value is determined such that the initial bone mask that over represents bone within the 3D image.

4. The method of claim 2 , wherein the global thresholding operation is a hysteresis thresholding operation that uses an upper threshold and a lower threshold determined using the global threshold value.

5. The method of claim 1 , comprising:
(n) following act (m), receiving, by the processor, via a graphical user interface (GUI), a user selection of one or more of a plurality of labeled regions of the labeled final bone map, wherein the user selection corresponds to an identification of the one or more labeled regions as corresponding to HO; and
(o) labeling, by the processor, the one or more labeled regions selected by the user as corresponding to HO and labeling, by the processor, the remaining labeled regions of the plurality of labeled regions as corresponding to normal skeleton, thereby producing a binary labeled normal skeleton and HO map that differentiates between regions of the 3D image corresponding to normal skeleton and regions of the image corresponding to HO.

6. The method of claim 5 , comprising determining, by the processor, one or more morphometric measurements using the binary labeled normal skeleton and HO map.

7. The method of claim 6 , comprising determining, by the processor, a total volume of the regions of the binary labeled normal skeleton and HO map that are labeled as corresponding to HO.

8. The method of claim 1 , wherein the one or more second derivative splitting filters comprises at least one member selected from the group consisting of a LoG (Laplacian of Gaussian), a HEH (highest Hessian eigenvalue, with preliminary Gaussian filtering), and a LEH (lowest Hessian eigenvalue, with preliminary Gaussian filtering).

9. The method of claim 1 , wherein applying the one or more second derivative splitting filters comprises applying a plurality of second derivative splitting filters, wherein applying the plurality of second derivative splitting filters comprises:
for each second derivative splitting filter being applied, producing a filtered image and identifying voxels of the filtered image with intensity higher or lower than a threshold value as split line voxels.

10. The method of claim 1 , comprising:
prior to applying the one or more second derivative splitting filters, performing, by the processor, a Gaussian filtering operation on the 3D image to produce a Gaussian filtered version of the 3D image; and
applying, by the processor, the one or more second derivative splitting filters to the Gaussian filtered version of the 3D image.

11. The method of claim 1 , wherein the 3D image of the subject is a CT image and wherein the method comprises acquiring the CT image.

12. A method for automatically detecting heterotopic ossification (HO) in a 3D image of a subject, the method comprising:
(a) receiving, by a processor of a computing device, the 3D image of a subject;
(b) applying, by the processor, a global thresholding operation to the received 3D image to produce an initial bone mask that identifies an initial region of interest within the 3D image comprising a graphical representation of bone;
(c) determining, by the processor, a boundary value map using a 3D edge detection operation applied to the initial region of interest of the 3D image identified by the initial bone mask, wherein the boundary value map identifies and includes intensity values of voxels of the 3D image that correspond to boundaries where bone meets soft tissue;
(d) determining, by the processor, a bone threshold map using the initial bone mask and the boundary value map, wherein the bone threshold map comprises, for each voxel of the initial bone mask, a threshold value determined by extrapolating values of the boundary value map to voxels within the initial bone mask;
(e) determining, by the processor, a final bone mask using the bone threshold map and the 3D image;
(f) applying, by the processor, one or more second derivative splitting filters to the 3D image to identify a set of split line voxels within the 3D image;
(g) removing, by the processor, voxels corresponding to the set of split line voxels from the final bone mask, thereby generating a first split bone mask;
(h) determining, by the processor, a distance map by applying a distance transform to the first split bone mask;
(i) applying, by the processor, a watershed segmentation operation to the distance map to identify at least one of a set of catchment basins and watershed lines within the distance map;
(j) generating, by the processor, a second split bone mask using (A) the first split bone mask and (B) at least one of the identified catchment basins and watershed lines from act (i);
(k) determining, by the processor, a plurality of labeled split binary components of the second split bone mask via one or more morphological processing operations;
(l) performing, by the processor, a region growing operation within the final bone mask using the plurality of lab

Assignees

• Perkinelmer Health Sci Inc

Inventors

• Behrooz Ali

Classifications

• G06T7/11 Primary

  Region-based segmentation · CPC title

• A61B6/5217 Primary

  extracting a diagnostic or physiological parameter from medical diagnostic data · CPC title

• G06V10/751

  Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching · CPC title

• G06T2207/30008

  Bone · CPC title

• G06T2207/20152

  Watershed segmentation · CPC title