Systems and methods for image correction in positron emission tomography

Patent metadata

Field	Value
Publication number	US-11501473-B2
Application number	US-201916728081-A
Country	US
Kind code	B2
Filing date	Dec 27, 2019
Priority date	Dec 28, 2018
Publication date	Nov 15, 2022
Grant date	Nov 15, 2022

Abstract

System for image correction in PET is provided. The system may acquire a PET image and a CT image of a subject. The system may generate, based on the PET image and the CT image, an attenuation-corrected PET image of the subject by application of an attenuation correction model. The attenuation correction model may be a trained cascaded neural network including a trained first model and at least one trained second model downstream to the trained first model. During the application of the attenuation correction model, an input of each of the at least one trained second model may include the PET image, the CT image, and an output image of a previous trained model that is upstream and connected to the trained second model.

First claim

What is claimed is:

1. A system for image correction in positron emission tomography (PET), 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:
acquiring a PET image and a computed tomography (CT) image of a subject;
generating a concatenated image based on the PET image and the CT image; and
generating, based on the concatenated image, an attenuation-corrected PET image of the PET image by application of an attenuation correction model, wherein:
the attenuation correction model is a trained cascaded neural network including a plurality of trained models that are sequentially connected,
the plurality of trained models include a trained first model and at least one trained second model downstream to the trained first model, and
during the application of the attenuation correction model, an input of each of the at least one trained second model includes the concatenated image and an output image of a previous trained model that is upstream and connected to the trained second model.

2. The system of claim 1 , wherein the generating a concatenated image based on the PET image and the CT image includes:
preprocessing the CT image and the PET image; and
generating the concatenated image by concatenating the preprocessed CT image and the preprocessed PET image.

3. The system of claim 2 , wherein the preprocessing the CT image and the PET image includes:
registering the CT image with the PET image;
generating a resampled CT image and a resampled PET image by resampling the registered CT image and the registered PET image, each of the resampled CT image and the resampled PET image having a preset image resolution; and
generating the preprocessed CT image and the preprocessed PET image by normalizing the resampled CT image and the resampled PET image.

4. The system of claim 3 , wherein the generating, based on the concatenated image, an attenuation-corrected PET image of the PET image by application of an attenuation correction model includes:
obtaining a preliminary attenuation-corrected PET image by inputting the concatenated image into the attenuation correction model;
denormalizing the preliminary attenuation-corrected PET image; and
generating the attenuation-corrected PET image by resampling the denormalized preliminary attenuation-corrected PET image, the attenuation-corrected PET image and the PET image having a same image resolution.

5. The system of claim 4 , wherein:
the attenuation correction model is trained using a plurality of sample attenuation-corrected PET images, and
the denormalization of the preliminary attenuation-corrected PET image is performed based on a mean value and a standard deviation of the plurality of sample attenuation-corrected PET images.

6. The system of claim 1 , wherein the acquiring a PET image and a CT image of a subject includes:
acquiring CT image data and PET image data of the subject by performing a CT scan and a PET scan of the subject;
reconstructing, based on the CT image data, the CT image;
reconstructing, based on the PET image data, a preliminary PET image; and
generating the PET image by performing a random correction and a detector normalization on the preliminary PET image.

7. The system of claim 1 , wherein at least one of the plurality of trained models is a convolutional neural network (CNN) model or a generative adversarial network (GAN) model.

8. The system of claim 1 , wherein the generation of the attenuation-corrected PET image of the PET image is performed within 1 second.

9. A system for generating an attenuation correction model, 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:
acquiring a plurality of training samples, each of the plurality of training samples including a sample positron-emission tomography (PET) image of a sample subject, a sample computed tomography (CT) image of the sample subject, and a sample attenuation-corrected PET image corresponding to the sample PET image;
for each of the plurality of training samples, generating a sample concatenated image based on the sample PET image and the sample CT image of the training sample; and
generating the attenuation correction model by training a cascaded neural network using the plurality of training samples and a plurality of sample concatenated images corresponding to the plurality of training samples, wherein:
the cascaded neural network includes a plurality of sequentially connected models,
the plurality of models includes a first model and at least one second model downstream to the first model, and
during the training of the cascaded neural network, each of the at least one second model is trained based on the plurality training samples and one or more models in the cascaded neural network upstream to the second model.

10. The system of claim 9 , wherein the plurality of models are trained in parallel during the training of the cascaded neural network, and the training the cascaded neural network includes:
initializing parameter values of the cascaded neural network; and
training the cascaded neural network by iteratively updating the parameter values of the cascaded neural network based on the plurality of training samples.

11. The system of claim 10 , wherein iteratively updating the parameter values of the cascaded neural network includes performing an iterative operation including one or more iterations, and each of at least one iteration of the iterative operation includes:
for each of at least some of the plurality of training samples, generating a predicted attenuation-corrected PET image by application of an updated cascaded neural network determined in a previous iteration;
determining, based on the predicted attenuation-corrected PET image and the sample attenuation-corrected PET image of each of the at least some of the plurality of training samples, an assessment result of the updated cascaded neural network; and
further updating the parameter values of the updated cascaded neural network to be used in a next iteration based on the assessment result, wherein during the application of the updated cascaded neural network to a training sample,
each second model of the updated cascaded neural network is configured to receive the training sample and an output image of a previous model that is upstream and connected to the second model in the updated cascaded neural network, and
the predicted attenuation-corrected PET image is an output image of a last second model of the sequentially connected models in the updated cascaded neural network.

12. The system of claim 11 , wherein the determining an assessment result of the updated cascaded neural network comprises:
for each of the plurality of models in the updated cascaded neural network, determining, based on the sample attenuation-corrected PET image and an output image of the model corresponding to each of the at least some of the plurality of training samples, a value of a loss function corresponding to the model; and
determining, based on the values of the loss functions of the plurality of models, the assessment result.

13. The system of claim 12 , wherein the parameter values of the cascaded neural network include pa

Assignees

• Shanghai United Imaging Intelligence Co Ltd

Inventors

• Zhang Yang
• Lyu Yang
• Xi Chen

Classifications

• G06T12/10 Primary

  Image preprocessing, e.g. calibration, positioning of sources or scatter correction · CPC title

• G06N3/084 Primary

  Backpropagation, e.g. using gradient descent · CPC title

• G06N3/047

  Probabilistic or stochastic networks · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06T12/30

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