US 12,013,451 B2
Noise adaptive data consistency in deep learning image reconstruction via norm ball projection
Simon Arberet, Princeton, NJ (US); Boris Mailhe, Plainsboro, NJ (US); Marcel Dominik Nickel, Herzogenaurach (DE); Thomas Benkert, Neunkirchen am Brand (DE); Mahmoud Mostapha, Princeton, NJ (US); and Mariappan S. Nadar, Plainsboro, NJ (US)
Assigned to Siemens Healthineers AG, Forchheim (DE)
Filed by Siemens Healthineers AG, Forchheim (DE)
Filed on Sep. 6, 2022, as Appl. No. 17/929,803.
Prior Publication US 2024/0077561 A1, Mar. 7, 2024
Int. Cl. G01R 33/48 (2006.01); G01R 33/54 (2006.01); G01R 33/565 (2006.01)
CPC G01R 33/4818 (2013.01) [G01R 33/546 (2013.01); G01R 33/565 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer implemented method for image reconstruction, the method comprising:
acquiring scan data defining an input image, wherein the scan data includes noise level information;
generating a reconstructed image from the scan data using a reconstruction network comprising at least a data consistency operation that enforces data consistency by using a norm ball projection that adjusts a balance between a reconstruction network prediction and the scan data based on at least the noise level information; and
outputting the reconstructed image.