Accelerating Fiber Tracking with CUDA
Author Information
Author(s): Evert van Aart, Neda Sepasian, Andrei Jalba, Anna Vilanova
Primary Institution: Eindhoven University of Technology
Hypothesis
Can the geodesic ray-tracing algorithm for fiber tracking be accelerated using CUDA on GPUs?
Conclusion
The CUDA implementation of the geodesic ray-tracing algorithm can achieve speedups of up to 40 times compared to traditional CPU implementations.
Supporting Evidence
- The CUDA implementation allows for a speedup of up to 40 times compared to CPU implementations.
- The algorithm can find multiple connections between regions in the brain.
- Using GPUs for fiber tracking reduces the computational time significantly.
Takeaway
This study shows how using powerful graphics cards can make tracking brain fibers much faster, helping doctors see important connections in the brain more quickly.
Methodology
The study implemented a geodesic ray-tracing algorithm on a GPU using CUDA, allowing for parallel processing of fiber tracking.
Limitations
The algorithm requires significant memory and has high computational complexity, which can limit performance.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website