Improving Arterial Spin Labeling using Deep Learning

Abstract

We proposed a new convolutional neural network (CNN) framework to quantify cerebral blood flow (CBF) in Hadamard-encoded pseudo-continuous arterial spin labeling (HE-pCASL). Improving sensitivity and robustness in ASL signals allows CNNs to quantify CBF accurately with a smaller number of data acquisitions. The proposed methods outperformed the conventional averaging method in both normal and pathologic regions. Therefore, CNNs can be a good alternative to quantify CBF in ASL imaging.