US 12,315,045 B2
Task-specific training of reconstruction neural network algorithm for magnetic resonance imaging reconstruction
Alexander Preuhs, Erlangen (DE); and Mario Orsini, Grand Rapids, MI (US)
Assigned to SIEMENS HEALTHINEERS AG, Forchheim (DE)
Filed by Siemens Healthcare GmbH, Erlangen (DE)
Filed on May 25, 2022, as Appl. No. 17/824,268.
Claims priority of application No. 21176482 (EP), filed on May 28, 2021.
Prior Publication US 2022/0392122 A1, Dec. 8, 2022
Int. Cl. G06T 11/00 (2006.01); G06V 10/82 (2022.01); G06V 20/50 (2022.01)
CPC G06T 11/005 (2013.01) [G06V 10/82 (2022.01); G06V 20/50 (2022.01); G06T 2210/41 (2013.01)] 18 Claims
OG exemplary drawing
 
1. A computer-implemented method of training a reconstruction neural network algorithm used to reconstruct a Magnetic Resonance Imaging (MRI) image based on MRI raw data, the computer-implemented method comprising:
obtaining training MRI raw data acquired using a k-space trajectory that is undersampling the k-space;
obtaining a ground truth of a training MRI image associated with the training MRI raw data;
obtaining a ground truth of a presence or absence of an object in the training MRI image;
obtaining a ground truth of a location of the object in the training MRI image;
determining a prediction of the training MRI image based on the training MRI raw data and using the reconstruction neural network algorithm;
determining a prediction of the presence or absence of the object in the training MRI image based on the prediction of the training MRI image and using an object-detection algorithm;
determining a prediction of the location of the object in the training MRI image based on the training MRI image and using the object-detection algorithm;
determining a loss value based on (i) a first difference between the ground truth of the training MRI image and the prediction of the training MRI image, and (ii) a second difference between the ground truth of the presence or absence of the object and the prediction of the presence or absence of the object, and (iii) a third difference between the ground truth of the location of the object and the prediction of the location of the object; and
adjusting weights of the reconstruction neural network algorithm based on the loss value and using a training process.