Graphics Processor Unit (GPU) Accelerated Shallow Transparent Layer Detection in Optical Coherence Tomographic (OCT) images for real-time Corneal Surgical Guidance
Abstract
An image analysis algorithm is described that utilizes a Graphics Processor Unit (GPU) to detect in real-time the most shallow subsurface tissue layer present in an OCT image obtained by a prototype SDOCT corneal imaging system. The system has a scanning depth range of 6mm and can acquire 15 volumes per second at the cost of lower resolution and signal-to-noise ratio (SNR) than diagnostic OCT scanners. To the best of our knowledge, we are the first to experiment with non-median percentile filtering for simultaneous noise reduction and feature enhancement in OCT images, and we believe we are the first to implement any form of non-median percentile filtering on a GPU. The algorithm was applied to five different test images. On an average, it took ~0.5 milliseconds to preprocess an image with a 20th-percentile filter, and ~1.7 milliseconds for our second-stage algorithm to detect the faintly imaged transparent surface.
BibTeX
@workshop{Mathai-2014-17167,author = {Tejas Sudharshan Mathai and John Galeotti and Samantha J. Horvath and George D. Stetten},
title = {Graphics Processor Unit (GPU) Accelerated Shallow Transparent Layer Detection in Optical Coherence Tomographic (OCT) images for real-time Corneal Surgical Guidance},
booktitle = {Proceedings of 9th International Workshop on Augmented Environments for Computer-Assisted Interventions (AE-CAI '14)},
year = {2014},
month = {September},
pages = {1 - 13},
keywords = {OCT, image-guidance, real-time, GPU, percentile filter, surface detection},
}