Graphics processing unit accelerated optical coherence tomography processing at megahertz axial scan rate and high...

Area of Science:

• Biomedical Imaging
• Computer Science
• Optical Engineering

Background:

• Optical Coherence Tomography (OCT) is a crucial imaging modality for real-time visualization.
• High-speed OCT systems generate large datasets requiring efficient processing.
• Current processing platforms often struggle to keep pace with advanced OCT acquisition rates.

Purpose of the Study:

• To develop and highly optimize a Computer Unified Device Architecture (CUDA)-based platform for real-time OCT data processing.
• To achieve high-speed three-dimensional (3-D) volumetric rendering of OCT data using a cost-effective Graphics Processing Unit (GPU).
• To demonstrate a processing rate capable of matching ultrahigh-speed OCT acquisition.

Main Methods:

• Utilized a commercially available, cost-effective Graphics Processing Unit (GPU) for computational acceleration.
• Implemented optimized algorithms for real-time processing of interferometric OCT data.
• Developed efficient 3-D volumetric rendering techniques, including B-scan and en face view display.

Main Results:

• Achieved a maximum axial scan processing rate of 2.24 MHz (16 bits pixel depth, 2048 FFT size), including memory transfer and B-scan display.
• Reached a maximum 3-D volumetric rendering rate of approximately 23 volumes/second (volume size: 1024×256×200).
• Reported the fastest processing rate to date using a single-chip GPU for OCT.

Conclusions:

• The optimized GPU platform enables real-time video-rate volumetric OCT processing and rendering.
• This implementation successfully matches the acquisition rates of ultrahigh-speed OCT systems.
• The cost-effective GPU approach offers a viable solution for high-performance OCT data analysis.