Sparsity transform k‐t principal component analysis for accelerating cine three‐dimensional flow measurements

Abstract

Time-resolved three-dimensional flow measurements are limited by long acquisition times. Among the various acceleration techniques available, k-t methods have shown potential as they permit significant scan time reduction even with a single receive coil by exploiting spatiotemporal correlations. In this work, an extension of k-t principal component analysis is proposed utilizing signal differences between the velocity encodings of three-directional flow measurements to further compact the signal representation and hence improve reconstruction accuracy. The effect of sparsity transform in k-t principal component analysis is demonstrated using simulated and measured data of the carotid bifurcation. Deploying sparsity transform for 8-fold undersampled simulated data, velocity root-mean-square errors were found to decrease by 52 ± 14%, 59 ± 11%, and 16 ± 32% in the common, external, and internal carotid artery, respectively. In vivo, errors were reduced by 15 ± 17% in the common carotid artery with sparsity transform. Based on these findings, spatial resolution of three-dimensional flow measurements was increased to 0.8 mm isotropic resolution with prospective 8-fold undersampling and sparsity transform k-t principal component analysis reconstruction. Volumetric data were acquired in 6 min. Pathline visualization revealed details of helical flow patterns partially hidden at lower spatial resolution.