Magnetic Resonance Spectroscopy Spectral Registration with Deep Learning

Abstract

Deep learning is an effective image processing approach that has been enthusiastically adopted in Magnetic Resonance Spectroscopy (MRS). Methods such as multilayer perceptrons (MLP) and convolutional neural networks (CNN) have been applied to frequency and phase correction (FPC) to help resolve frequency and phase shifts that arise in MRS. However, both methods need to be trained separately with frequency and phase offsets to perform FPC. In this study, we aim to introduce a spectrum registration technique using CNNs that perform simultaneous correction of both frequency and phase shifts of single voxel MEGA-PRESS MRS simulated data.