US 11,720,794 B2
Training a multi-stage network to reconstruct MR images
Bryan Clifford, Malden, MA (US); Thorsten Feiweier, Poxdorf (DE); Steffen Bollman, Gehofen (DE); and Stephen Farman Cauley, Somerville, MA (US)
Assigned to Siemens Healthcare GmbH, Erlangen (DE); and The General Hospital Corporation, Boston, MA (US)
Filed by Siemens Healthcare GmbH, Erlangen (DE); and The General Hospital Corporation, Boston, MA (US)
Filed on Feb. 18, 2021, as Appl. No. 17/178,674.
Prior Publication US 2022/0261629 A1, Aug. 18, 2022
Int. Cl. G06N 3/08 (2023.01); G06N 3/045 (2023.01); G06T 11/00 (2006.01)
CPC G06N 3/08 (2013.01) [G06N 3/045 (2023.01); G06T 11/003 (2013.01); G06T 2210/41 (2013.01); G06T 2211/424 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A system comprising:
a storage device storing a plurality of target examples, each of the plurality of target examples including respective target imaging data and target acquisition parameters; and
a processing unit to execute processor-executable program code to cause the system to:
generate, for each target example, modified imaging data and modified acquisition parameters of a training example based on the target imaging data and target acquisition parameters of the target example;
generate, for each training example, an initial reconstructed image based on the modified imaging data and modified acquisition parameters of the training example;
train a first network stage of a multi-stage network based on the modified imaging data, modified acquisition parameters and initial reconstructed image of each training example;
generate a first output image for each training example by inputting the modified imaging data, modified acquisition parameters and initial reconstructed image of the training example to the trained first network stage; and
train a second network stage of the multi-stage network based on the modified imaging data, modified acquisition parameters and first output image of each training example.
 
8. A method comprising:
generating, for each target example, modified imaging data and modified acquisition parameters of a training example based on the target imaging data and target acquisition parameters of the target example;
generating, for each training example, an initial reconstructed image based on the modified imaging data and modified acquisition parameters of the training example;
training a first network stage of a multi-stage network based on the modified imaging data, modified acquisition parameters and initial reconstructed image of each training example, and on a target image generated based on the target imaging data and target acquisition parameters of each target example;
generating a first output image for each training example by inputting the modified imaging data, modified acquisition parameters and initial reconstructed image of the training example to the trained first network stage; and
train a second network stage of the multi-stage network based on the modified imaging data, modified acquisition parameters and first output image of each training example, and on the target image generated based on the target imaging data and target acquisition parameters of each target example.
 
15. A system comprising:
a training example generator to:
generate modified imaging data and modified acquisition parameters of training examples based on target imaging data and target acquisition parameters of each of a plurality of target examples; and
generate an initial reconstructed image based on the modified imaging data and modified acquisition parameters of each training example;
a training system to:
train a first network stage of a multi-stage network based on the modified imaging data, modified acquisition parameters and initial reconstructed image of each training example, and on a target image associated with a corresponding target example;
generate a first output image for each training example by inputting the modified imaging data, modified acquisition parameters and initial reconstructed image of the training example to the trained first network stage; and
train a second network stage of the multi-stage network based on the modified imaging data, modified acquisition parameters and first output image of each training example, and on a target image associated with a corresponding target example.