Improved BPE reconstruction using FOCUSS

Introduction: Bunched Phase Encoding (BPE) is a new type of fast data acquisition method in MRI that takes advantage of zigzag k-space trajectories along the PE direction (Fig.1)[1,2]. Sampling frequency of BPE is higher than that of normal acquisition. Since BPE acquisition scheme is comparable to acquiring multiple PE lines in a single readout, the total number of TR cycles and hence the scan time can be reduced. A primary disadvantage of BPE is that images reconstructed using matrix inversion methods are sometimes affected by high levels of noise. Various efforts have been made to reduce noise and aliasing artifacts by modifying reconstruction methods. For example, recent studies of parallel imaging have demonstrated that regularization methods are useful to mitigate noise amplification [3-5]. However, these methods often may result in loss of spatial resolution or exhibit residual aliasing artifacts. The focal underdetermined system solver (FOCUSS) has recently been successfully adapted to undersampled projection reconstruction, spiral and dynamic imaging in MRI [6-8]. FOCUSS capitalizes on sparsity of image support and considerably reduces aliasing artifacts from undersampled k-space data. In this study, we present a novel framework to reduce SNR loss in BPE reconstruction. In our newly proposed technique, high frequency k-space data are processed using a combined method of regularization and FOCUSS to reconstruct images with reduced noise and aliasing artifacts. The newly proposed method is referred to as ‘BPE-FOCUSS’. Images reconstructed using BPE-FOCUSS show substantially reduced noise from those reconstructed using a conventional BPE reconstruction method while no apparent aliasing artifacts are introduced. BPE-FOCUSS is quite useful and facilitates implementation of BPE in practice.