US 11,990,224 B2
Synthetically generating medical images using deep convolutional generative adversarial networks
Hamid Jafarkhani, Irvine, CA (US); Saeed Karimi-Bidhendi, Irvine, CA (US); and Arash Kheradvar, Irvine, CA (US)
Assigned to THE REGENTS OF THE UNIVERSITY OF CALIFORNIA, Oakland, CA (US)
Filed by THE REGENTS OF THE UNIVERSITY OF CALIFORNIA, Oakland, CA (US)
Filed on Mar. 26, 2021, as Appl. No. 17/214,442.
Claims priority of provisional application 63/000,401, filed on Mar. 26, 2020.
Prior Publication US 2021/0312242 A1, Oct. 7, 2021
Int. Cl. G06T 7/143 (2017.01); G06F 18/21 (2023.01); G06F 18/214 (2023.01); G06N 3/08 (2023.01); G06N 7/01 (2023.01); G06T 7/00 (2017.01); G06V 10/82 (2022.01); G16H 30/40 (2018.01)
CPC G16H 30/40 (2018.01) [G06F 18/2148 (2023.01); G06F 18/2185 (2023.01); G06F 18/2193 (2023.01); G06N 3/08 (2013.01); G06N 7/01 (2023.01); G06T 7/0012 (2013.01); G06T 7/143 (2017.01); G06V 10/82 (2022.01); G06T 2207/10088 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30048 (2013.01); G06V 2201/031 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A generative adversarial network stored on a non-transitory computer readable medium, comprising:
a generator configured to generate synthetic medical image data attributed to a cardiovascular system and segmentation masks corresponding to the synthetic medical image data, the generator having parameters that have been initialized according to a predetermined probability distribution; and
a discriminator configured to receive the synthetic medical image data from the generator and determine probabilities indicating likelihood of the synthetic medical image data corresponding to real cardiovascular images acquired from an individual, the discriminator having parameters that have been initialized according to the predetermined probability distribution, wherein the discriminator is further configured to provide the probabilities determined by the discriminator to: (1) the generator and (2) an input of the discriminator to allow the parameters of the generator and the parameters of the discriminator to be iteratively adjusted until an equilibrium between the generator and the discriminator is established,
wherein the generator is configured to output two channels comprising a first channel that outputs information representing the synthetic medical image data and a second channel that outputs information representing the segmentation masks corresponding to the synthetic medical image data.