Back to results
US 12,599,351 B2Granted
Imaging having scatter correction by interpolation
AssigneeSiemens Healthineers AG
InventorPhilipp Roser
Life of this patent
Estimating the term…
Abstract
In order to reduce a computing effort for correcting initial images from an imaging apparatus, a scatter model is used to simulate scattered-radiation images, and corresponding initial images are selected from the original initial images. These are used to train a neural network. The neural network may be used to calculate scattered-radiation single images that may be used to correct each individual initial image into a corrected initial image.
Claims15
A method for training a neural network for correcting acquisitions obtained by an imaging apparatus, the method comprising:
- acquiring a plurality of initial images at different acquisition coordinates using the imaging apparatus;
- obtaining a number of simulated scattered-radiation images from the plurality of initial images using a scatter model, the number of simulated scattered-radiation images being less than the plurality of initial images, wherein each simulated scattered-radiation image of the number of simulated scattered-radiation images is assigned one corresponding acquisition coordinate;
- selecting images of the plurality of initial images that have respective acquisition coordinates equal to acquisition coordinates of the number of simulated scattered-radiation images;
- and
- training the neural network using the selected images as input data and the number of simulated scattered-radiation images as output data.
2 more elements
A method for correcting initial images obtained by radiation, the method comprising:
- creating one scattered-radiation single image for each of the initial images using a neural network trained according to a method for training the neural network for correcting acquisitions obtained by an imaging apparatus, the method for training the neural network comprising:
- acquiring a plurality of initial images at different acquisition coordinates using the imaging apparatus;
- obtaining a number of simulated scattered-radiation images from the plurality of initial images using a scatter model, the number of simulated scattered-radiation images being less than the plurality of initial images, wherein each simulated scattered-radiation image of the number of simulated scattered-radiation images is assigned one corresponding acquisition coordinate;
- selecting images of the plurality of initial images that have respective acquisition coordinates equal to acquisition coordinates of the number of simulated scattered-radiation images;
- and
- training the neural network using the selected images as input data and the number of simulated scattered-radiation images as output data;
- and
- correcting each of the initial images using the corresponding scattered-radiation single image.
5 more elements
An imaging system comprising:
- an imaging apparatus;
- and
- an image processing apparatus configured to train a neural network for correcting acquisitions obtained by the imaging apparatus, the image processing apparatus being configured to train the neural network comprising the image processing apparatus being configured to:
- acquire a plurality of initial images at different acquisition coordinates using the imaging apparatus;
- obtain a number of simulated scattered-radiation images from the plurality of initial images using a scatter model, the number of simulated scattered-radiation images being less than the plurality of initial images, wherein each simulated scattered-radiation image of the number of simulated scattered-radiation images is assigned one corresponding acquisition coordinate;
- select images of the plurality of initial images that have respective acquisition coordinates equal to acquisition coordinates of the number of simulated scattered-radiation images;
- and
- train the neural network using the selected images as input data and the number of simulated scattered-radiation images as output data.
5 more elements
In a non-transitory computer-readable storage medium executable by a control facility of an imaging system to correct initial images obtained by radiation, the instructions comprising:
- creating one scattered-radiation single image for each of the initial images using a neural network trained according to a method for training the neural network for correcting acquisitions obtained by an imaging apparatus, the method for training the neural network comprising:
- acquiring a plurality of initial images at different acquisition coordinates using the imaging apparatus;
- obtaining a number of simulated scattered-radiation images from the plurality of initial images using a scatter model, the number of simulated scattered-radiation images being less than the plurality of initial images, wherein each simulated scattered-radiation image of the number of simulated scattered-radiation images is assigned one corresponding acquisition coordinate;
- selecting images of the plurality of initial images that have respective acquisition coordinates equal to acquisition coordinates of the number of simulated scattered-radiation images;
- and
- training the neural network using the selected images as input data and the number of simulated scattered-radiation images as output data;
- and
- correcting each of the initial images using the corresponding scattered-radiation single image.
5 more elements
Dependencies are inferred from the claim text.