Improved R2* and QSM mapping for dummies - ask Adam

In this abstract we demonstrate a simple framework that builds up on deep learning infrastructure to perform quantitative susceptibility mapping and correct for susceptibility related effects on R2* maps, taking the advantage of high GPU computational efficiency. Asking the Adam optimizer to point you in the right direction results in a gradient descent method that can perform field mapping, background field removal, QSM and reduce macroscopic intravoxel dephasing artifacts in R2* maps with most operations being performed in under 60 seconds even for 0.8mm isotropic whole brain multi-echo data.