Comparison of real-time visualization of volumetric OCT data sets by CPU-slicing and GPU-ray casting methods Alfred R. Fuller a , Robert J. Zawadzki b* , Bernd Hamann a and John S. Werner b a Visualization and Computer Graphics Research Group, Institute for Data Analysis and Visualization (IDAV), Department of Computer Science, UC Davis, One Shields Avenue, Davis, CA 95616, USA; b Vision Science and Advanced Retinal Imaging Laboratory (VSRI) and Department of Ophthalmology & Vision Science, UC Davis, 4860 Y Street, Suite 2400, Sacramento, CA 95817, USA; ABSTRACT We describe and compare two volume visualization methods for Optical Coherence Tomography (OCT) retinal data sets. One of these methods is CPU-slicing, which is previously reported and used in our visualization engine. The other is GPU-ray casting. Several metrics including image quality, performance, hardware limitations and perception are used to grade the abilities of each method. We also discuss how to combine these methods to make a scalable volume visualization system that supports advanced lighting and dynamic volumetric shadowing techniques on a broad range of hardware. The feasibility of each visualization method for clinical application as well as potential further improvements are discussed. Keywords: Optical coherence tomography, imaging system, medical optics instrumentation, ophthalmology, volume visualization 1. INTRODUCTION Recent progress in Fourier domain-Optical Coherence Tomography (Fd-OCT) 1–5 has allowed for successful im- plementation for ophthalmic clinical applications. 6–8 The fast acquisition speeds of Fd-OCT have led to the volumetric imaging of in vivo retinal structures resulting in large volumetric data sets. Thus there is grow- ing demand for volume visualization and manipulation software. Following this direction over last three years our group has developed custom volume visualization software. 9 This software has proven to be a powerful tool for conveying the information acquired using Optical Coherence Tomography (OCT) through computer visualization. The goal of computer visualization is to convey meaningful information about a data set through a visual representation. This is done by simulating stimuli on a computer screen that the human brain is accustomed to interpreting. However, computer visualization is limited by the resources of the computer generating the stimuli. This means that it can be costly in terms of computer processing power and resources to produce physically accurate stimuli akin to the natural world we experience and interpret every day. By dividing the perceptual experience into specific visual cues and targeting the specific perceptions associated with these cues, computer visualization has been able to find a middle ground between limited computer resources and the desired conveyance of information. Naturally, as computers advance, this middle ground is continually being pushed forward to produce more meaningful visualizations. Volume visualization of OCT retinal data sets strives to convey composition, shape, structure, and relative depth and size of various retinal structures. The basic method behind volume visualization uses a transfer function to convert the data values to visual stimuli. Visual stimuli are generated via color and opacity values which are projected onto a computer screen to convey the basic composition of the data set. Additional visual cues are added to further enhance the user’s perception of the data set. *rjzawadzki@ucdavis.edu; phone 1 916 734-5839; fax 1 916 734-4543; http://vsri.ucdavis.edu/