Anatomy-aware motion estimation
AssigneeShanghai United Imaging Intelligence Co., Ltd.
InventorsXiao Chen, Pingjun Chen, Zhang Chen, Terrence Chen, Shanhui Sun
Life of this patent
Estimating the term…
Abstract
Described herein are neural network-based systems, methods and instrumentalities associated with estimating the motion of an anatomical structure. The motion estimation may be performed utilizing pre-learned knowledge of the anatomy of the anatomical structure. The anatomical knowledge may be learned via a variational autoencoder, which may then be used to optimize the parameters of a motion estimation neural network system such that, when performing motion estimation for the anatomical structure, the motion estimation neural network system may produce results that conform with the underlying anatomy of anatomical structure.
Claims20
A neural network system comprising one or more processors and one or more storage devices, the one or more storage devices configured to store instructions that, when executed by the one or more processors, cause the one or more processors to:
- receive a first medical image and a second medical image, wherein the first medical image comprises a first visual representation of an anatomical structure and the second medical image comprises a second visual representation of the anatomical structure;
- extract a first plurality of features from the first medical image and a second plurality of features from the second medical image;
- determine a first motion of the anatomical structure between the first medical image and the second medical image based on the first plurality of features and the second plurality of features respectively extracted from the first medical image and the second medical image;
- and
- generate a first flow field indicating the first motion;
- wherein the neural network system is trained at least partially using a variational autoencoder (VAE) as a supervisor of the training, the VAE being pre-trained to generate a refined segmentation mask for the anatomical structure based on a shape prior of the anatomical structure, and wherein, during the training of the neural network system, the neural network system is configured to generate a coarse segmentation mask for the anatomical structure based on a first plurality of features extracted from a first training image or based on a second plurality of features extracted from a second training image, and adjust one or more operating parameters of the neural network system based on a first loss calculated from the coarse segmentation mask and the refined segmentation mask.
A method for motion estimation implemented by a neural network system, the method comprising:
- receiving a first medical image and a second medical image, wherein the first medical image comprises a first visual representation of an anatomical structure and the second medical image comprises a second visual representation of the anatomical structure;
- extracting a first plurality of features from the first medical image and a second plurality of features from the second medical image;
- determining a first motion of the anatomical structure between the first medical image and the second medical image based on the first plurality of features and the second plurality of features respectively extracted from the first medical image and the second medical image;
- and
- generating a first flow field indicating the first motion;
- wherein the neural network system is trained at least partially using a variational autoencoder (VAE) as a supervisor of the training, the VAE being pre-trained to generate a refined segmentation mask for the anatomical structure based on a shape prior of the anatomical structure, and wherein, during the training of the neural network system, the neural network system is configured to generate a coarse segmentation mask for the anatomical structure based on a first plurality of features extracted from a first training medical image or based on a second plurality of features extracted from a second training medical image, and adjust one or more operating parameters of the neural network system based on a first loss calculated from the coarse segmentation mask and the refined segmentation mask.
Dependencies are inferred from the claim text.