Multi-GPU Reconstruction of Dynamic Compressed Sensing MRI
Magnetic resonance imaging (MRI) is a widely used in-vivo imaging technique that is essential to the diagnosis of disease, but its longer acquisition time hinders its wide adaptation in time-critical applications, such as emergency diagnosis. Recent advan
- PDF / 614,931 Bytes
- 9 Pages / 439.363 x 666.131 pts Page_size
- 18 Downloads / 243 Views
Abstract. Magnetic resonance imaging (MRI) is a widely used in-vivo imaging technique that is essential to the diagnosis of disease, but its longer acquisition time hinders its wide adaptation in time-critical applications, such as emergency diagnosis. Recent advances in compressed sensing (CS) research have provided promising theoretical insights to accelerate the MRI acquisition process, but CS reconstruction also poses computational challenges that make MRI less practical. In this paper, we introduce a fast, scalable parallel CS-MRI reconstruction method that runs on graphics processing unit (GPU) cluster systems for dynamic contrast-enhanced (DCE) MRI. We propose a modified Split-Bregman iteration using a variable splitting method for CS-based DCE-MRI. We also propose a parallel GPU Split-Bregman solver that scales well across multiple GPUs to handle large data size. We demonstrate the validity of the proposed method on several synthetic and real DCE-MRI datasets and compare with existing methods.
1
Introduction
Magnetic resonance imaging (MRI) has been widely used as an in-vivo imaging technique due to its safety to living organisms. Because the acquisition process is the major bottleneck of MRI, many acceleration techniques have been developed using parallel imaging techniques [2]. Recently, the compressed sensing (CS) theory [3] has been successfully adopted to MRI to speed up the acquisition process [10]. However, CS-MRI introduces additional computational overhead in the reconstruction process because the 1 minimization is a time-consuming nonlinear optimization problem. Therefore, there exists a need to develop fast CS reconstruction methods to make the entire MRI reconstruction process practical for time-critical applications. One research direction for accelerating CS reconstruction has focused on developing efficient numerical solvers for the 1 minimization problem [7,5,1]. The other direction has been leveraging the state-of-the-art parallel computing hardware, such as the graphics processing unit (GPU), to push the performance to the limit [9,13,1]. We believe that GPU acceleration is the most promising approach to make CS-MRI reconstruction clinically feasible, but multi-GPU acceleration has not been fully addressed in previously published literature. ∗
Corresponding author.
c Springer International Publishing Switzerland 2015 N. Navab et al. (Eds.): MICCAI 2015, Part III, LNCS 9351, pp. 484–492, 2015. DOI: 10.1007/978-3-319-24574-4_58
Multi-GPU Reconstruction of Dynamic Compressed Sensing MRI
(a) ×8 subsampled k-space
(b) Zero padding recon.
485
(c) CS-DCE-MRI recon.
Fig. 1. CS-DCE-MRI examples. Red line: time axis, Blue–green lines: x–y axis
In this paper, we introduce a novel multi-GPU reconstruction method for CS-based dynamic contrast-enhanced (DCE) MRI. DCE-MRI is widely used to detect tumors, and its diagnostic accuracy highly depends on the spatio-temporal resolution of data. Fig. 1 shows an example of CS-based DCE-MRI reconstruction from a sub-sampled k-space data. CS has been s
Data Loading...
