FIGURE 1.

Overview of the pipeline for simulating the optimal patch design based on pulmonary artery CT. The GAN comprises a generator CNN G and a discriminator CNN D. Regression is used to determine the pulmonary artery value at every voxel in a pulmonary artery stenosis CT. This is done by a skip connection which augments an estimated patch from hemodynamics analysis to the input stenosis CT image. The discriminator is trained to discriminate repaired CT images from normal pulmonary artery CT images.