Acceleration of fiber tracking in DTI tractography by reconfigurable computer hardware

Diffusion Tensor Imaging (DTI) tractography is a computationally intensive procedure. The most time consuming operation is the tracking of fibers from every voxel in the scanned volume. Fiber tracking can be accelerated significantly by use of reconfigurable hardware, such as Field Programmable Gate Arrays (FPGAs), which can track fibers at very high speed by exploiting the flexibility, parallelism and high on-chip bandwidth. Such acceleration has the potential to lead to realtime tractography. In this paper we isolate key kernels within the tracking step and through a simulation study, analyze a specific FPGA architecture comprising deeply pipelined kernel chains running in parallel. Our results suggest that the FPGA based computer architecture could achieve a two orders of magnitude speed-up in the fiber-tracking algorithm over an optimized C-code.