A novel compressed sensing inspired approach for flow reconstruction

Abstract

High scan times are one of the most important drawbacks in 4D flow scans and multiple solutions have been proposed to solve this issue. We propose a novel method for undersampled flow reconstruction inspired on the ideas of compressed sensing. By considering the magnitude and complex phase as separate variables, we were able to impose independent properties on each, such as having a constant magnitude over all flow enconding acquisitions and enforcing low phase values on low magnitude areas, thus directly reducing the resulting images’ noise. Our method was able to successfully reconstruct flow data with negligible error from undersampled data.