查看文章

Background: Quantification of aortic flow is increasingly utilized for patient management. While 4D Flow MRI provides a comprehensive assessment of aortic hemodynamics, it requires long acquisition times and is not widely available. Alternatively, contrast enhanced MR angiography (CEMRA) is easily acquired, becoming clinical routine, but provides only anatomical information. Deriving hemodynamics directly from CEMRA could significantly impact patient management. Seeking to address this using artificial intelligence (AI), we developed a CycleGAN for the prediction of systolic 3D blood flow velocity vector fields directly from CEMRA images.

Methods: This study used a total of paired 771 CEMRA and aortic 4D flow datasets (median age: 42 years, 556M/215F), acquired during the same exam on either 1.5 T or 3T MRI systems (Siemens). All scans were acquired with 3D coverage of the thoracic aorta with …