AI032
Programming Massively Parallel Processors: A Hands-on Approach
Case Study: Parallelizing MRI Reconstruction
Learning Objectives
- Identify the computational bottlenecks in traditional serial MRI reconstruction pipelines.
- Evaluate the performance gains achieved by parallelizing gridding algorithms for non-Cartesian k-space data.
- Compare the efficiency of OpenMP and CUDA implementations for 3D Fast Fourier Transforms.
- Analyze the impact of data-level parallelism on the convergence speed of iterative SENSE and GRAPPA reconstruction.