QR-decomposition based SENSE reconstruction using parallel architecture.

Area of Science:

• Medical Imaging
• Computational Imaging
• Biomedical Engineering

Background:

• Magnetic Resonance Imaging (MRI) is crucial for clinical diagnosis but suffers from long scan times.
• Parallel Magnetic Resonance Imaging (pMRI) algorithms, like Sensitivity Encoding (SENSE), can accelerate MRI acquisition.
• SENSE reconstruction involves computationally intensive matrix inversion, often a bottleneck.

Purpose of the Study:

• To develop and evaluate a novel Graphics Processing Unit (GPU) based implementation of the SENSE algorithm.
• To assess the performance of the GPU-SENSE algorithm against single and multicore Central Processing Unit (CPU) implementations.
• To demonstrate the potential of GPU acceleration for reducing MRI scan times.

Main Methods:

• A novel GPU-based SENSE algorithm was implemented utilizing QR decomposition for encoding matrix inversion.
• Performance was evaluated by comparing computation times against single and multicore CPU implementations using OpenMP.
• Experiments were conducted using multichannel phantom, human head, and cardiac datasets at various acceleration factors.

Main Results:

• The GPU-based SENSE implementation achieved significant speedups compared to CPU-based methods.
• Approximately 12x speedup was observed against multicore CPU and 53x speedup against single-core CPU.
• Image quality of the reconstructed MR images remained undegraded.

Conclusions:

• GPU acceleration offers a substantial reduction in computation time for SENSE-based MRI reconstruction.
• The proposed GPU-SENSE method effectively accelerates MRI scans without compromising diagnostic image quality.
• This advancement holds promise for improving the efficiency and accessibility of advanced MRI techniques.