Improving MRI Reconstruction with Compressive Sensing
Author Information
Author(s): Kim Daehyun, Trzasko Joshua, Smelyanskiy Mikhail, Haider Clifton, Dubey Pradeep, Manduca Armando
Primary Institution: Mayo Clinic
Hypothesis
Can advanced computing architectures significantly improve the performance of compressive sensing MRI reconstructions?
Conclusion
The study demonstrates that optimized use of modern many-core architectures can significantly reduce the computational time required for compressive sensing MRI reconstructions, making them more clinically viable.
Supporting Evidence
- The CUDA-based code on an NVIDIA Tesla C2050 GPU can reconstruct a 256 × 160 × 80 volume in 19 seconds.
- Intel's Knights Ferry can perform the same reconstruction in only 12 seconds.
- The optimized implementation on a dual-socket six-core CPU can reconstruct the same volume in 35 seconds.
Takeaway
This study shows that using powerful computers can help doctors get better MRI images faster, which is really important for helping patients.
Methodology
The study investigates different throughput-oriented architectures for compressive sensing MRI reconstruction and compares their performance.
Limitations
The study primarily focuses on specific architectures and may not generalize to all MRI applications.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website