Systems and methods for deep learning-based image reconstruction

Patent metadata

Field	Value
Publication number	US-11227418-B2
Application number	US-201816235046-A
Country	US
Kind code	B2
Filing date	Dec 28, 2018
Priority date	Dec 28, 2018
Publication date	Jan 18, 2022
Grant date	Jan 18, 2022

Abstract

Methods, apparatus and systems for deep learning based image reconstruction are disclosed herein. An example at least one computer-readable storage medium includes instructions that, when executed, cause at least one processor to at least: obtain a plurality of two-dimensional (2D) tomosynthesis projection images of an organ by rotating an x-ray emitter to a plurality of orientations relative to the organ and emitting a first level of x-ray energization from the emitter for each projection image of the plurality of 2D tomosynthesis projection images; reconstruct a three-dimensional (3D) volume of the organ from the plurality of 2D tomosynthesis projection images; obtain an x-ray image of the organ with a second level of x-ray energization; generate a synthetic 2D image generation algorithm from the reconstructed 3D volume based on a similarity metric between the synthetic 2D image and the x-ray image; and deploy a model instantiating the synthetic 2D image generation algorithm.

First claim

What is claimed is:

1. At least one non-transitory computer-readable storage medium including instructions that, when executed, cause at least one processor to at least:
obtain a plurality of two-dimensional (2D) tomosynthesis projection images of an organ by rotating an x-ray emitter to a plurality of orientations relative to the organ and emitting a first level of x-ray energization from the emitter for each projection image of the plurality of 2D tomosynthesis projection images;
reconstruct a three-dimensional (3D) volume of the organ from the plurality of 2D tomosynthesis projection images;
obtain an x-ray image of the organ with a second level of x-ray energization;
generate a synthetic 2D image generation learning algorithm from the plurality of two-dimensional (2D) tomosynthesis projection images and the reconstructed 3D volume based on a similarity metric between the synthetic 2D image and the x-ray image; and
deploy an artificial intelligence model instantiating the synthetic 2D image generation learning algorithm to form a synthetic image.

2. The at least one computer-readable storage medium of claim 1 , wherein the x-ray image is to be registered to fit in a geometry of the plurality of 2D tomosynthesis projection images.

3. The at least one computer-readable storage medium of claim 1 , wherein the plurality of 2D tomosynthesis projection images of the organ and the x-ray image of the organ are to be obtained during a same compression of the organ.

4. The at least one computer-readable storage medium of claim 1 , wherein the plurality of 2D tomosynthesis projection images and the x-ray image are further to be obtained with a detector that receives x-rays emitted from the x-ray emitter, the instructions, when executed, to further cause the at least one processor to:
applying a dynamic range correction factor to at least one of the plurality of 2D projection images or the x-ray image.

5. The at least one computer-readable storage medium of claim 1 , wherein each plane of the reconstructed 3D volume is to match with a geometry of the x-ray image.

6. The at least one computer-readable storage medium of claim 1 , wherein the instructions, when executed, further cause the at least one processor to:
map each pixel of the synthetic 2D image to at least one voxel in the reconstructed 3D volume;
present the synthetic 2D image in a graphical user interface (GUI) generated on a graphical display;
receive a user selection of an object of interest in the x-ray image;
identify at least one plane through the 3D volume; and
present the at least one identified plane on the graphical display.

7. The at least one computer-readable storage medium of claim 1 , wherein the instructions, when executed, further cause the at least one processor to:
enhance the synthetic 2D image with the areas extracted from the reconstructed 3D volume; and
enrich a mapping of the synthetic 2D image to the reconstructed 3D volume with the location of the extracted areas.

8. The at least one computer-readable storage medium of claim 1 , wherein the instructions, when executed, further cause the at least one processor to:
enhance the x-ray image with the areas extracted from the reconstructed 3D volume,
wherein the synthetic 2D image generation learning algorithm is generated from the reconstructed 3D volume based on a similarity metric between the synthetic 2D image and the enhanced x-ray image.

9. The at least one computer-readable storage medium of claim 1 , wherein an energy to obtain the x-ray image is higher than an energy to obtain the plurality of two-dimensional (2D) tomosynthesis projection images.

10. The at least one computer-readable storage medium of claim 1 , wherein generating the synthetic 2D image generation learning algorithm includes determining the synthetic 2D image generation learning algorithm using a training model such that the synthetic 2D image generation learning algorithm tends to minimize the similarity metric between the synthetic 2D image and the x-ray image.

11. The at least one computer-readable storage medium of claim 1 , wherein the model includes an artificial neural network model.

12. The at least one computer-readable storage medium of claim 8 , wherein the areas extracted from the reconstructed 3D volume are provided by at least one of an operator or a computer-aided detection system.

Assignees

• Gen Electric

Inventors

• Bernard Sylvain

Classifications

• G06T12/30 Primary

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

• G06T2200/24

  involving graphical user interfaces [GUIs] · CPC title

• G06T2207/10072

  Tomographic images · CPC title

• G06T2211/436

  Limited angle · CPC title

• G06T2210/41

  Medical · CPC title