Parallelized multi-graphics processing unit framework for high-speed Gabor-domain optical coherence microscopy.

Area of Science:

• Biomedical Optics
• Medical Imaging Technology
• Computational Imaging

Background:

• Gabor-domain optical coherence microscopy (GD-OCM) offers high-resolution 3-D skin imaging.
• Clinical applications of GD-OCM are limited by the need for real-time image processing.

Purpose of the Study:

• To develop and evaluate a parallelized, multi-graphics processing unit (GPU) computing framework for real-time GD-OCM image processing.
• To achieve computational speeds that exceed GD-OCM acquisition rates for clinical usability.

Main Methods:

• Implemented a parallelized control mechanism to distribute computation tasks across multiple GPUs.
• Optimized the number of amplitude-scans (A-scans) processed per GPU to maximize memory usage and core throughput.
• Investigated five computing architectures for computational speed-up on 1000x1000 A-scan datasets.

Main Results:

• The multi-GPU framework achieved processing speeds faster than GD-OCM acquisition.
• Processing a 1x1x0.6 mm³ skin sample took approximately 13 seconds with two GPUs and 6.5 seconds with four GPUs.
• Demonstrated real-time display of 3-D GD-OCM data is feasible.

Conclusions:

• The proposed parallelized multi-GPU framework significantly accelerates GD-OCM image processing.
• This advancement facilitates high-speed, real-time 3-D skin imaging for clinical applications.
• Enables immediate visualization of histological-resolution skin data for clinicians.