Swept-source endoscopic optical coherence tomography real-time imaging system based on GPU acceleration for axial...

Area of Science:

• Biomedical Optics
• Medical Imaging Technology
• High-Speed Data Processing

Background:

• Real-time image processing is crucial for high-speed endoscopic swept-source optical coherence tomography (SS-OCT).
• Existing systems face challenges in processing speed and image drift at high acquisition rates.
• Optimized platforms are needed to enhance the diagnostic capabilities of SS-OCT.

Purpose of the Study:

• To develop and optimize a computer-unified device architecture (CUDA)-based platform for real-time image processing and correction in high-speed SS-OCT.
• To improve processing throughput and suppress image drift for enhanced endoscopic imaging.
• To establish a foundation for non-destructive, in vivo, early tumor diagnosis using SS-OCT.

Main Methods:

• Implemented a Half-Sync/Half-Asyn memory optimization mode to create a high-throughput data thread pool for the CPU.
• Utilized an asynchronous streaming architecture for high-speed thread multiplexing to accelerate data processing.
• Designed a rotary scanning position information encoding feedback module with FPGA synchronization control to mitigate image drift.

Main Results:

• Achieved a maximum complete axial-scan-processing rate of 3.52 MHz (16-bit pixel depth, 1024 A-scans/s).
• Reported the fastest processing rate to date for SS-OCT using a single-chip graphical processing unit.
• Demonstrated stable real-time imaging of mouse esophagus and human fingers at 200 frames per second (1024x1024 pixels).

Conclusions:

• Developed a real-time image processing and reconstruction technology for high-throughput SS-OCT systems.
• The technology enables high-density operation, efficient parallelism, and suppression of high-speed image drift.
• Lays the groundwork for fast, convenient, non-destructive, in vivo, non-staining early tumor diagnosis with endoscopic SS-OCT.