Tomographic image reconstruction using artificial intelligence (AI) engines

Patent metadata

Field	Value
Publication number	US-11436766-B2
Application number	US-201916722004-A
Country	US
Kind code	B2
Filing date	Dec 20, 2019
Priority date	Dec 20, 2019
Publication date	Sep 6, 2022
Grant date	Sep 6, 2022

Abstract

Example methods and systems for tomographic image reconstruction are provided. One example method may comprise: obtaining two-dimensional (2D) projection data and processing the 2D projection data using the AI engine that includes multiple first processing layers, an interposing back-projection module and multiple second processing layers. Example processing using the AI engine may involve: generating 2D feature data by processing the 2D projection data using the multiple first processing layers, reconstructing first three-dimensional (3D) feature volume data from the 2D feature data using the back-projection module; and generating second 3D feature volume data by processing the first 3D feature volume data using the multiple second processing layers.

First claim

We claim:

1. A method for a computer system to perform tomographic image reconstruction using an artificial intelligence (AI) engine, wherein the method comprises:
obtaining two-dimensional (2D) projection data associated with a target object and acquired using an imaging system;
processing the 2D projection data using the AI engine that includes multiple first processing layers, an interposing back-projection module and multiple second processing layers by performing the following:
generating 2D feature data by processing the 2D projection data using the multiple first processing layers that are associated with first weight data and operate in a 2D projection space;
reconstructing first three-dimensional (3D) feature volume data from the 2D feature data using the back-projection module; and
generating second 3D feature volume data by processing the first 3D feature volume data using the multiple second processing layers that are associated with second weight data and operate in a 3D volume space.

2. The method of claim 1 , wherein generating the second 3D feature volume data comprises:
generating the second 3D feature volume data by processing the first 3D feature volume data using the multiple second processing layers that are trained to perform encoding on the first 3D feature volume data.

3. The method of claim 1 , wherein generating the 2D feature data comprises:
generating the 2D feature data by processing the 2D projection data using the multiple first processing layers that are trained to perform pre-processing on the 2D projection data in the 2D projection space.

4. The method of claim 3 , wherein generating the 2D feature data comprises:
generating the 2D feature data by processing the 2D projection data using the multiple first processing layers that are trained to apply one or more convolution filters on the 2D projection data.

5. The method of claim 1 , wherein the method further comprises:
obtaining training data that includes (a) training 2D projection data and (b) training 3D feature volume data; and
training the multiple first processing layers and the multiple second processing layers together, with the back-projection module interposed in between, to transform (a) the training 2D projection data and (b) the training 3D feature volume data and to learn the respective first weight data and second weight data.

6. The method of claim 1 , wherein the method further comprises:
processing the second 3D feature volume data generated by the AI engine, being a first AI engine, using a second AI engine that is trained to perform tomographic image analysis to transform the second 3D feature volume data to analysis output data.

7. The method of claim 6 , wherein the method further comprises:
obtaining training data that includes (a) training 2D projection data and (b) training analysis output data; and
training the first AI engine and the second AI engine together to perform integrated tomographic image reconstruction and analysis to transform (a) the training 2D projection data to (b) the training analysis output data.

8. A non-transitory computer-readable storage medium that includes a set of instructions which, in response to execution by a processor of a computer system, cause the processor to perform a method of tomographic image reconstruction using an artificial intelligence (AI) engine, wherein the method comprises:
obtaining two-dimensional (2D) projection data associated with a target object and acquired using an imaging system;
processing the 2D projection data using the AI engine that includes multiple first processing layers, an interposing back-projection module and multiple second processing layers by performing the following:
generating 2D feature data by processing the 2D projection data using the multiple first processing layers that are associated with first weight data and operate in a 2D projection space;
reconstructing first three-dimensional (3D) feature volume data from the 2D feature data using the back-projection module; and
generating second 3D feature volume data by processing the first 3D feature volume data using the multiple second processing layers that are associated with second weight data and operate in a 3D volume space.

9. The non-transitory computer-readable storage medium of claim 8 , wherein generating the second 3D feature volume data comprises:
generating the second 3D feature volume data by processing the first 3D feature volume data using the multiple second processing layers that are trained to perform encoding on the first 3D feature volume data.

10. The non-transitory computer-readable storage medium of claim 8 , wherein generating the 2D feature data comprises:
generating the 2D feature data by processing the 2D projection data using the multiple first processing layers that are trained to perform pre-processing on the 2D projection data in the 2D projection space.

11. The non-transitory computer-readable storage medium of claim 10 , wherein generating the 2D feature data comprises:
generating the 2D feature data by processing the 2D projection data using the multiple first processing layers that are trained to apply one or more convolution filters on the 2D projection data.

12. The non-transitory computer-readable storage medium of claim 8 , wherein the method further comprises:
obtaining training data that includes (a) training 2D projection data and (b) training 3D feature volume data; and
training the multiple first processing layers and the multiple second processing layers together, with the back-projection module interposed in between, to transform (a) the training 2D projection data and (b) the training 3D feature volume data and to learn the respective first weight data and second weight data.

13. The non-transitory computer-readable storage medium of claim 8 , wherein the method further comprises:
processing the second 3D feature volume data generated by the AI engine, being a first AI engine, using a second AI engine that is trained to perform tomographic image analysis to transform the second 3D feature volume data to analysis output data.

14. The non-transitory computer-readable storage medium of claim 13 , wherein the method further comprises:
obtaining training data that includes (a) training 2D projection data and (b) training analysis output data; and
training the first AI engine and the second AI engine together to perform integrated tomographic image reconstruction and analysis to transform (a) the training 2D projection data to (b) the training analysis output data.

15. A computer system configured to perform tomographic image reconstruction using an artificial intelligence (AI) engine, wherein the computer system comprises: a processor and a non-transitory computer-readable medium having stored thereon instructions that, when executed by the processor, cause the processor to:
obtain two-dimensional (2D) projection data associated with a target object and acquired using an imaging system;
process the 2D projection data using the AI engine that includes multiple first processing layers, an interposing back-projection module and multiple second processing layers by performing the following:
generate 2D feature data by processing the 2D projection data using the multiple first processing layers that are associated with first weight data and operate in a 2D projection space;
reconstruct first three-dimensional (3D) feature volume data from the 2D feature data using the back-projection module; and
generate second 3D feature volume data by processing the first 3D feature volume data using the multiple secon

Assignees

• Varian Medical Systems Int Ag

Inventors

• Paysan Pascal
• Haas Benjamin M
• Nord Janne
• Perttu Sami Petri
• Seghers Dieter
• Pyyry Joakim

Classifications

• G06T12/20 Primary

  Inverse problem, i.e. transformations from projection space into object space · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title