Refining lesion contours with combined active contour and inpainting

Patent metadata

Field	Value
Publication number	US-11587236-B2
Application number	US-202017085042-A
Country	US
Kind code	B2
Filing date	Oct 30, 2020
Priority date	Oct 30, 2020
Publication date	Feb 21, 2023
Grant date	Feb 21, 2023

Abstract

A mechanism is provided in a data processing system for refining lesion contours with combined active contour and inpainting. The mechanism receives an initial segmented medical image having organ tissue including a set of object contours and a contour to be refined. The mechanism inpaints object voxels inside all contours of the set. The mechanism calculates an updated contour around the contour to be refined based on the in-painted object voxels to form an updated segmented medical image. The mechanism determines whether the updated segmented medical image is improved compared to the initial segmented medical image. The mechanism keeps the updated segmented medical image responsive to the updated segmented medical image being improved.

First claim

What is claimed is:

1. A method, in a data processing system comprising at least one processor and at least one memory, the at least one memory comprising instructions executed by the at least one processor to implement lesion segmentation logic for refining lesion contours with combined active contour and inpainting, the method comprising:
receiving an initial segmented medical image having organ tissue including a set of first object contours and a second object contour that is to be refined;
inpainting object voxels inside all first contours of the set of first object contours such that voxels inside the first contours and voxels in the medical image that are not within the second object contour have a same voxel value;
calculating an updated second object contour based on voxel values of the in-painted object voxels inside all first contours and voxels in the medical image that are not within the second object contour, to form an updated segmented medical image;
determining whether the updated second object contour of the updated segmented medical image is improved compared to the second object contour in the initial segmented medical image; and
keeping the updated segmented medical image having the updated second object contour, responsive to the updated segmented medical image being improved, wherein the initial segmented medical image comprises a plurality of lesion contours, and wherein the method is repeated for each lesion contour in the plurality of lesion contours.

2. The method of claim 1 , further comprising:
keeping the initial segmented medical image responsive to updated segmented medical image not being improved.

3. The method of claim 1 , wherein determining whether the updated segmented medical image is improved comprises:
determining a first contrast for the initial segmented medical image;
determining a second contrast for the updated segmented medical image; and
determining whether the second contrast are greater than the first contrast.

4. The method of claim 1 , wherein determining whether the updated segmented medical image is improved comprises:
determining a first variance between voxels inside each contour and voxels outside the contour but remaining within a predetermined distance away from the contour for the initial segmented medical image;
determining a second variance between voxels inside each contour and voxels outside the contour but remaining within a pre-determined distance away from the contour for the updated segmented medical image; and
determining whether the second variance is less than the first variance.

5. The method of claim 4 , wherein determining the first variance and the second variance comprises calculating variance of a set of n voxels {x 1 , . . . , x n } as follows:

∑

i
=
1

n

x
i
2

n

-

(

∑

i
=
1

n

x
i

n

)

2

.

6. The method of claim 1 , wherein the method is repeated for each lesion contour in the plurality of lesion contours at least by, for each repeat of the method, selecting a different lesion contour from a previously selected lesion contour, in the plurality of lesion contours, to be the second object contour, and including other lesion contours in the plurality of lesion contours to be part of the set of first object contours.

7. The method of claim 1 , wherein inpainting object voxels inside all first contours of the set of first object contours such that voxels inside the first contours and voxels in the medical image that are not within the second object contour have a same voxel value comprises setting voxel values for the voxels inside all first contours and voxels in the medical image that are not within the second object contour to an average tissue value of voxels not within the second object contour.

8. The method of claim 1 , wherein the set of object contours comprises at least one of a contour of a non-organ part of the segtr ented medical image, a contour of a blood vessel, or a contour of an artery.

9. The method of claim 1 , wherein, the second contour corresponds to a lesion, and wherein inpainting the object voxels inside the set of first object contours comprises inpainting non-organ tissue voxels in a vicinity of the lesion.

10. The method of claim 1 , wherein the lesion segmentation logic is a stage of computer logic in a lesion detection and classification artificial intelligence (AI) pipeline comprising a plurality of trained machine learning computer models, and wherein the lesion segmentation logic receives input from a liver/lesion detection logic stage of the AI pipeline.

11. The method of claim 10 , wherein the plurality of trained machine learning computer models of the AI pipeline comprises:
one or more first machine learning computer models of the AI pipeline that process an input volume of medical images to detect lesions present in the medical images of the input volume which correspond to an anatomical structure of interest; and
one or more second machine learning computer models of the AI pipeline that processes the detected lesions to perform lesion segmentation and combining of lesion contours of different medical images in the input medical image associated with a same lesion, to generate the listing of lesions and corresponding lesion contours.

12. The method of claim 10 , wherein the plurality of trained machine learning computer models comprise one or more machine learning computer models that process the updated segmented medical image output by the lesion segmentation logic to remove false positives from the updated segmented medical image.

13. A computer program product comprising a computer readable storage medium having a computer readable program stored therein, wherein the computer readable program, when executed on a computing device, causes the computing device to implement lesion segmentation logic for refining lesion contours with combined active contour and inpainting, wherein the lesion segmentation logic executes to:
receive an initial segmented medical image having organ tissue including a set of first object contours and a second object contour that is to be refined;
inpaint object voxels inside all first contours of the set of first object contours such that voxels inside the first contours and voxels in the medical image that are not within the second object contour have a same voxel value;
calculate an updated second object contour based on voxel values of the in-painted object voxels inside all first contours and voxels in the medical image

Assignees

• IBM

Inventors

• Palma Giovanni John Jacques
• Wang yi-qing

Classifications

• G06T12/30

  Image post-processing, e.g. metal artefact correction · CPC title

• A61B6/50

  specially adapted for specific body parts; specially adapted for specific clinical applications · CPC title

• G06T2207/20084

  Artificial neural networks [ANN] · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06T2207/30056

  Liver; Hepatic · CPC title