Item

Abstract

An end-to-end differentiable Bloch
simulation is employed to reduce T2 induced
blurring in single-shot RARE sequences, by
using a joint optimization of refocusing flip
angles and a convolutional neural network. The
instantaneous transverse magnetization,
available in our simulation, at a certain echo
time, serves as ideal blurring-free target. The
optimized RARE was able to successfully lower
T2-induced blurring for single-shot RARE
sequences in PDw and T2w images. In addition
to an increased sharpness, the neural network
allowed correction of the contrast changes to
match the theoretical transversal
magnetization.