Method and apparatus for training generative model for 3D consistent medical volume data
AssigneeGengenai, Inc
InventorsHojin Cho, Sangil Kim
Life of this patent
Estimating the term…
Abstract
A method for training a 3D space consistent medical volume data generative model is provided, which is performed by one or more processors and which includes generating a synthetic volume data using a generator network, receiving real volume data, generating a plurality of 2D images based on the synthetic volume data and the real volume data, discriminating whether each of the plurality of generated 2D images is real or fake by using a plurality of discriminator networks, and updating the generator network based on discrimination results of the plurality of discriminator networks.
Claims20
A method for training a 3D space consistent medical volume data generative model, the method being performed by one or more processors and comprising:
- receiving real volume data that comprises three-dimensional scan data of at least a portion of a human;
- generating a first plurality of 2D slice images by projecting at least a portion of the real volume data into a first plane;
- receiving, via a generator network and based on random input to the generator network, generator network output comprising synthetic volume data;
- generating, based on the synthetic volume data, a second plurality of 2D slice images by projecting the synthetic volume data in the first plane;
- selecting, based on the first plane, a first discriminator network of a plurality of discriminator networks;
- providing, to the first discriminator network, input data comprising the first plurality of 2D slice images and the second plurality of 2D slice images;
- receiving, from the first discriminator network, output indicating for each 2D slice image in the input data, whether the 2D slice image is real or fake;
- and
- updating, based on the output from the first discriminator network, the generator network.
An apparatus comprising:
- a memory;
- and
- one or more processors connected to the memory and configured to execute one or more computer-readable programs included in the memory,
- wherein the one or more computer-readable programs comprise instructions that when executed by the one or more processors, cause:
- receiving real volume data that comprises three-dimensional scan data of at least a portion of a human;
- generating a first plurality of 2D slice images by projecting at least a portion of the real volume data into a first plane;
- receiving, via a generator network and based on random input to the generator network, generator network output comprising synthetic volume data;
- receiving real volume data;
- generating, based on the synthetic volume data, a second plurality of 2D slice images based on the synthetic volume data and the real volume data by projecting the synthetic volume data in the first plane;
- selecting, based on the first plane, a first discriminator network of a plurality of discriminator networks;
- providing, to the first discriminator network, input data comprising the first plurality of 2D slice images and the second plurality of 2D slice images;
- receiving, from the first discriminator network, output indicating, for each 2D slice image in the input data, whether the 2D slice image is real or fake;
- and
- updating, based on the output from the first discriminator network, the generator network.
A system for generating medical volume data, comprising:
- a generator network;
- a plurality of discriminator networks;
- and
- a computing device configured to:
- receive real volume data that comprises three-dimensional scan data of at least a portion of a human;
- generate a first plurality of 2D slice images by projecting at least a portion of the real volume data into a first plane;
- receive, from the generator network and based on random input to the generator network, generator network output comprising synthetic volume data;
- generate, based on the synthetic volume data, a second plurality of 2D slice images by projecting the synthetic volume data in the first plane;
- select, based on the first plane, a first discriminator network of the plurality of discriminator networks;
- provide, to the first discriminator network, input data comprising the first plurality of 2D slice images and the second plurality of 2D slice images;
- receive, from the first discriminator network, output indicating, for each 2D slice image in the input data, whether the 2D slice image is real or fake;
- and
- update, based on the output from the first discriminator network, the generator network.
Dependencies are inferred from the claim text.