3-D brain MRI tissue classification on FPGAs.

Area of Science:

• Medical Imaging
• Computational Neuroscience
• Computer Engineering

Background:

• Magnetic Resonance Imaging (MRI) analysis algorithms are crucial for brain mapping.
• Increasingly large MRI datasets necessitate faster computational methods.
• Partial Volume Estimation (PVE) is a key brain tissue classification technique in MRI.

Purpose of the Study:

• To accelerate the Partial Volume Estimation (PVE) algorithm for MRI data analysis.
• To implement and evaluate PVE on a Field-Programmable Gate Array (FPGA) platform.
• To extend prior work by including probability density estimation acceleration.

Main Methods:

• Implementation of PVE, including probability density and prior information estimation, on an FPGA using Mitrion-C.
• Evaluation using simulated and real human brain MRI datasets.
• Performance comparison against an Itanium 2 CPU.

Main Results:

• FPGA implementation achieved average speedups of 2.5x for probability density estimation and 9.4x for prior information estimation.
• Overall performance improvement for the FPGA-based PVE algorithm reached 5.1x using four FPGAs.
• Demonstrated accuracy and significant performance gains for FPGA-accelerated MRI analysis.

Conclusions:

• FPGA-based acceleration offers substantial performance improvements for MRI PVE algorithms.
• This approach is effective for handling large-scale brain mapping datasets.
• The implemented FPGA solution enhances computational efficiency in neuroimaging analysis.