Granted Patent B2

US 12,586,161 · App. 18/056,778 · Granted Mar 24, 2026

Inventors: Wenjing Cao (Shanghai, CN); Haohua Sun (Shanghai, CN); Liyi Zhao (Shanghai, CN)

Assignee: SHANGHAI UNITED IMAGING HEALTHCARE CO., LTD.

G06T5/70G06T11/005G06T11/006G06T11/008G06T2211/424G06T2211/436

Abstract

Claims (73)

1 . A method for image reconstruction, which is implemented on a computing device including at least one processor and a computer-readable storage device, the method comprising:

obtaining an initial image to be processed; and

generating a reconstructed image by performing a plurality of iteration steps on the initial image, one of the plurality of iteration steps including a first optimization operation and a second optimization operation, wherein

the first optimization operation includes:

receiving an image to be processed in the iteration step, wherein the image to be processed in the iteration step includes the initial image, or an output of an optimizing model in a previously adjacent iteration step; and

determining an updated image by preliminarily optimizing the image to be processed, wherein the first optimization operation includes an iterative operation, the determining an updated image by preliminarily optimizing the image to be processed includes:

determining the updated image by optimizing, according to a first loss function, the image to be processed according to the iterative operation until a value of the first loss function satisfies a termination condition;

the second optimization operation includes:

determining, using the optimizing model, an optimized image based on the updated image; and

designating the optimized image as a next image to be processed in a next iteration step or designating the optimized image as the reconstructed image.

2 . The method of claim 1 , wherein the preliminarily optimizing the image to be processed includes:

optimizing the image to be processed according to the first loss function relating to a difference between forward projection data associated with the image to be processed and originally acquired projection data associated with the initial image.

3 . The method of claim 2 , wherein the first loss function relates to at least one of a first quality weight associated with a quality of the originally acquired projection data or a second quality weight associated with a quality of the image to be processed.

4 . The method of claim 3 , wherein the originally acquired projection data is acquired by at least one detector row of a radiation imaging device, and the first loss function relates to a third weight associated with a cone angle corresponding to the at least one detector row.

5 . The method of claim 2 , wherein the first loss function includes at least one of a classical regularization term or a machine learning regularization term that involves a machine learning model.

6 . The method of claim 5 , wherein a first ratio of the classical regularization term to the machine learning regularization term in a first iteration step is different from a second ratio of the classical regularization term to the machine learning regularization term in a second iteration step.

7 . The method of claim 1 , wherein the determining, using an optimizing model, an optimized image based on the updated image includes at least one of:

reducing noise information of the updated image using the optimizing model,

reducing artifact information of the updated image using the optimizing model,

improving a gray distribution of the updated image using the optimizing model,

improving a global gray scale of the updated image using the optimizing model,

improving a resolution of the updated image using the optimizing model,

improving the contrast of the updated image using the optimizing model, or

enhancing the updated image using the optimizing model.

8 . The method of claim 1 , further comprising:

determining, using at least one optimizing model, the optimized image based on the updated image each of the at least one optimizing model corresponding to one of different optimizing goals.

9 . The method of claim 8 , wherein the at least one optimizing model is applied in series or in parallel to optimize the updated image.

10 . The method of claim 1 , further including:

determining a quality feature of the updated image, wherein the determining, using an optimizing model, an optimized image based on the updated image further includes inputting the quality feature and the updated image into the optimizing model.

11 . The method of claim 10 , wherein the quality feature includes at least one of a noise feature, an artifact feature, a gray distribution, a global gray scale, a resolution, or a contrast of the updated image.

12 . The method of claim 1 , wherein the optimizing model is obtained by a training process including:

obtaining a plurality of training samples, each of the plurality of training samples including a gold standard image and a sample image, wherein the sample image is determined based on the gold standard image, and the gold standard image refers to an image without interference information or an image with interference information less than a threshold; and

obtaining the optimizing model by training a preliminary optimizing model based on the plurality of training samples.

13 . The method of claim 12 , wherein the obtaining a plurality of training samples includes:

obtaining a reference interference component;

obtaining a gold standard image; and

for each of the plurality of training samples, determining a sample image of the training sample by adding, according to an interference level, the reference interference component to the gold standard image.

14 . The method of claim 13 , wherein the obtaining a reference interference component includes:

obtaining the gold standard image corresponding to a first imaging dose;

obtaining a second reconstructed image corresponding to a second imaging dose lower than the first imaging dose; and

determining the reference interference component based on the gold standard image and the second reconstructed image.

15 . The method of claim 1 , wherein the designating the optimized image as a next image to be processed in a next iteration step or designating the optimized image as the reconstructed image includes:

determining whether the optimized image satisfies the termination condition in the iteration step;

in response to determining that the optimized image satisfies the termination condition, designating the optimized image as the reconstructed image; and

in response to determining that the optimized image does not satisfy the termination condition, designating the optimized image as the next image to be processed in the next iteration step.

16 . A system for image reconstruction, comprising:

at least one storage device including a set of instructions; and

at least one processor configured to communicate with the at least one storage device, wherein when executing the set of instructions, the at least one processor is configured to direct the system to perform operations including:

obtaining an initial image to be processed; and

generating a reconstructed image by performing a plurality of iteration steps on the initial image, one of the plurality of iteration steps including a first optimization operation and a second optimization operation, wherein

the first optimization operation includes:

receiving an image to be processed in the iteration step, wherein the image to be processed in the iteration step includes the initial image, or an output of an optimizing model in a previously adjacent iteration step; and

determining an updated image by preliminarily optimizing the image to be processed, wherein the first optimization operation includes an iterative operation, the determining an updated image by preliminarily optimizing the image to be processed includes:

determining the updated image by optimizing, according to a first loss function, the image to be processed according to the iterative operation until a value of the first loss function satisfies a termination condition;

the second optimization operation includes:

determining, using an optimizing model, an optimized image based on the updated image; and

designating the optimized image as a next image to be processed in a next iteration step or designating the optimized image as the reconstructed image.

17 . A non-transitory computer readable medium, comprising executable instructions that, when executed by at least one processor, direct the at least one processor to perform a method, the method comprising:

obtaining an initial image to be processed; and

generating a reconstructed image by performing a plurality of iteration steps on the initial image, one of the plurality of iteration steps including a first optimization operation and a second optimization operation, wherein

the first optimization operation includes:

receiving an image to be processed in the iteration step, wherein the image to be processed in the iteration step includes the initial image, or an output of an optimizing model in a previously adjacent iteration step; and

determining an updated image by preliminarily optimizing the image to be processed, wherein the first optimization operation includes an iterative operation, the determining an updated image by preliminarily optimizing the image to be processed includes:

determining the updated image by optimizing, according to a first loss function, the image to be processed according to the iterative operation until a value of the first loss function satisfies a termination condition:

the second optimization operation includes:

determining, using an optimizing model, an optimized image based on the updated image; and

designating the optimized image as a next image to be processed in a next iteration step or designating the optimized image as the reconstructed image.

18 . The method of claim 3 , wherein the first quality weight is determined based on at least one of interference information in originally acquired projection data associated with the initial image, parameters for acquiring the originally acquired projection data, or a signal-to-noise ratio of the originally acquired projection data.

19 . The method of claim 1 , wherein the iterative operation includes multiple iterations, and each iteration includes:

determining forward projection data by performing a forward projection transformation on the image to be processed in the iteration step;

determining a weighted error between the forward projection data and originally acquired projection data associated with the initial image based on at least one of a first quality weight, a second quality weight, or a third quality weight, the forward projection data, and the originally acquired projection data; and

determining the updated image based at least in part on back projection data of the weighted error between the forward projection data and the originally acquired projection data associated with the initial image.

20 . The method of claim 1 , wherein the first loss function includes a fidelity term and one or more regularization terms, the fidelity term and the one or more regularization terms have a same input of the image to be processed.

Assignments (1)

Priority Claims (1)

Continuity (2)

References Cited (69)