Undersampling to accelerate time-resolved MRI velocity measurement of carotid blood flow

Abstract

Time-resolved velocity information of carotid blood flow can be used to estimate haemodynamic conditions associated with carotid artery disease leading to stroke. MRI provides high-resolution measurement of such information but long scan time limits its clinical application in this area. In order to reduce scan time the MRI signal is often undersampled by skipping part of the signal during data acquisition. The aim of this work is to implement and evaluate different undersampling techniques for carotid velocity measurement on a 1.5 T clinical scanner. Most recent undersampling techniques assume spatial and temporal redundancies of real time-resolved MRI signal. In these techniques different undersampling strategies were proposed. Prior information or different assumptions of the nature of true signal were used in signal reconstruction. A brief review of these techniques and details of a representative technique, known as k-t BLAST, are presented. Another undersampling scheme, termed ktVD, is proposed to use predesigned undersampling patterns with variable sampling densities in both temporal and spatial dimensions. It aims to collect enough signal content at the signal acquisition stage and simplify signal reconstruction. Fidelity of the results from undersampled data is affected by many factors, such as signal dynamic content, degree of signal redundancy, noise level, degree of undersampling, undersampling patterns, and parameters of post-processing algorithms. Simulations and in vivo scans were conducted to investigate the effects of these factors in time-resolved 2D scans and time-resolved 3D scans. The results suggested velocity measurement became less reliable when they were obtained from less than 25% of the full signal. In time-resolved 3D scans the signal can be undersampled in either one or two spatial dimensions in addition to the temporal dimension. This allows more options in the design of undersampling patterns, which were tested in vivo. In order to test undersampling in three dimensions in high resolution 3D scans and measure velocity in three dimensions, a flow phantom was also scanned at high degrees of undersampling to test the proposed method.