Magnetic Resonance Image Processing Based on FPGAs

Magnetic resonance imaging (MRI) plays an important role for medical imaging technology in brain anatomical studies. Many automatic algorithms have been proposed for analyzing magnetic resonance imaging data sets. With the increasingly large data sets being used in mapping, there are some limitations of the manual algorithms, such as long computation time and issues regarding reproducibility. There has been a significant rise in the need for accelerating these algorithms. This paper handled the partial volume estimation (PVE), a brain tissue classification algorithm for MRI, on a field-programmable gate array (FPGA) as high performance reconfigurable computer using the Mitrion-C high-level language (HLL). FPGAs are very flexible and could achieve significant performance improvement by exploiting parallelism in processing. We used several simulated and real human brain MRI to evaluate the performance improvement of the proposed algorithm. The performance of probability density and prior information estimation implementation based on FPGAs achieved the overall speedup.