Purpose
To develop a “Compute Unified Device Architecture” (CUDA) based real-time (RT) graphics processor unit (GPU) visualization system that can display three dimensional (3D) triangle meshes of platform independent triangle mesh segmented optical coherence tomography (OCT) data sets in both red/cyan anaglyph as well as polarized stereo, in order to increase the clarity of subtle structures on the vitreo-retinal interface.
Methods
OCT platform independent 3D OCT data of the vitreo-retinal interface was collected on devices from four different OCT manufacturers. This data was segmented and meshed into triangle meshes using visualization methods developed for Maxon Cinema 4D™ (C4D) and Nextlimit Realflow™ (RF). These meshes were then imported into C4D and rendered in stereoscopic RT using OTOY Octane Renderer™ (OR). The data was rendered directly on a pair of Nvidia GeForce GTX 590™ GPUs providing 1536MB of memory per GPU and a total of 2048 CUDA processing cores. The stereoscopic visualization was achieved by viewing the data using polarized or anaglyph glasses.
Results
It was possible to create stereoscopic RT 3D OCT visualizations with ambient occlusion (AO) as well “High Dynamic Range Image” (HDRI) environment lighting at a speed of 28.325 Ms/sec at over 2 million Tris. Anisotropic specularity and variable index refraction with translucency was included.
Conclusions
Although the speed and the quality of the RT 3D OCT was high and the resulting visualizations of the vitreo-retinal interface increased the clarity of subtle structures, the actual pre-visualization data processing is complicated, time consuming and inefficient. Therefore this method remains interesting for scientific and educational purposes but needs to be streamlined before becoming clinically useful.
Keywords: 688 retina •
549 image processing •
550 imaging/image analysis: clinical