Graphics Process Units Research Articles

Real-time en-face Gabor optical coherence tomographic angiography on human skin using CUDA GPU.

We recently proposed an optical coherence tomographic angiography (OCTA) algorithm, Gabor optical coherence tomographic angiography (GOCTA), which can extract microvascular signals from a spectral domain directly with lower computational complexity compared to other algorithms. In this manuscript, we combine a programmable swept source, an OCT complex signal detecting unit, and graphics process units (GPU) to achieve a real-time en-face GOCTA system for human skin microvascular imaging. The programmable swept source can balance the A-scan rate and the spectral tuning range; the polarization-modulation based complex signal detecting unit can double the imaging depth range, and the GPU can accelerate data processing. C++ and CUDA are used as the programming platform where five parallel threads are created for galvo-driving signal generation, data acquisition, data transfer, data processing, and image display, respectively. Two queues (for the raw data and en-face images, respectively) are used to improve the data exchange efficiency among different devices. In this study, the data acquisition time and data processing time for each 3D complex volume (256×304×608 pixels,) are 405.3 and 173.7 milliseconds respectively. To the best of our knowledge, this is the first time to show en-face microvascular images covering 3×3 mm2 at a refresh rate of 2.5 Hz.

Graphics process unit accelerated lattice Boltzmann simulation of indoor air flow: Effects of sub-grid scale model in large-eddy simulation

In this present study, three-dimensional lattice Boltzmann method is implemented with the popular turbulence modeling method large-eddy simulation incorporating three different non-dynamic sub-grid scale models Smagorinsky, Vreman, and wall-adapting local eddy-viscosity for finding the inhomogeneous turbulent airflow patterns inside a model room with a partition. The large eddy simulation-lattice Boltzmann method code is validated with the experimental results of Posner’s model, where the model room having one partition at the bottom, one inlet, an outlet placed at top wall considered for the comparisons. The lattice Boltzmann method code is also validated without any sub-grid scale model with the results of lid-driven flow in a cubic cavity. The present numerical simulations are performed by the graphics process unit accelerated parallel programs using compute unified device architecture C platform. Double precession capable a Tesla k40 with 2880 compute unified device architecture cores NVIDIA graphics process unit card has been used for these simulations. Graphics processor units have gained popularity in recent years as a propitious platform for numerical simulation of fluid dynamics. In fact, faster computational task performance in graphics process units is one of the key factors for researchers to choose graphics process unit over conventional central processing units for the implementation of data-intensive numerical methods like lattice Boltzmann method. The effects of the sub-grid scale model have been evaluated in terms of the mean velocity profiles, streamlines as well as turbulence characteristics and found that there are significant differences in the results due to the different sub-grid scale models.

GPU-accelerated large eddy simulation of stirred tanks

A fast and economical solver, accelerated by the Graphics Process Units (GPU) of a single graphics card in a desktop computer, is developed for the simulation of stirred tanks, integrating Lattice Boltzmann Method (LBM), coherent Large Eddy Simulation (LES) and Immersed Boundary Method (IBM). The grid resolution can reach 13.8 million nodes in maximum, resolving considerable details of flow field in stirred tanks with only such a simple desktop computer. In the meantime, the computational speed is 1500-fold faster than that of the traditional transient simulation of Computational Fluid Dynamics (CFD) based on Navier-Stokes equations implemented on 16 CPU cores. It only takes less than two minutes to update the simulation of one impeller revolution, enabling the transient simulation of longer physical time of stirred tanks. We find that at least 4000 impeller revolutions in simulation is needed to achieve an obviously time-independent macro-instability frequency. With the coherent LES model and reasonable grid resolution, the fast solver is able to resolve more than 95% of total Turbulent Kinetic Energy (TKE) for highly turbulent region, reproducing the monotonic decrease of TKE along the radial direction, much better than the Smagorinsky model. The average velocity and TKE predicted by the fast solver are in better agreement with the experimental data in literature. The solver is therefore more suitable for the fast simulation of stirred tanks using only a desktop computer without the need of much finer grid resolution and supercomputers.

Multimodality Neurological Data Visualization With Multi-VOI-Based DTI Fiber Dynamic Integration.

Brain lesions are usually located adjacent to critical spinal structures, so it is a challenging task for neurosurgeons to precisely plan a surgical procedure without damaging healthy tissues and nerves. The advancement of medical imaging technologies produces a large amount of neurological data, which are capable of showing a wide variety of brain properties. Advanced algorithms of medical data computing and visualization are critically helpful in efficiently utilizing the acquired data for disease diagnosis and brain function and structure exploration, which is helpful for treatment planning. In this paper, we describe new algorithms and a software framework for multiple volume of interest specified diffusion tensor imaging (DTI) fiber dynamic visualization. The displayed results have been integrated with a volume rendering pipeline for multimodality neurological data exploration. A depth texture indexing algorithm is used to detect DTI fiber tracts in graphics process units (GPUs), which makes fibers to be displayed and interactively manipulated with brain data acquired from functional magnetic resonance imaging, T1- and T2-weighted anatomic imaging, and angiographic imaging. The developed software platform is built on an object-oriented structure, which is transparent and extensible. It provides a comprehensive human-computer interface for data exploration and information extraction. The GPU-accelerated high-performance computing kernels have been implemented to enable our software to dynamically visualize neurological data. The developed techniques will be useful in computer-aided neurological disease diagnosis, brain structure exploration, and general cognitive neuroscience.

Visualizing 3D/4D environmental data using many-core graphics processing units (GPUs) and multi-core central processing units (CPUs)

Visualizing 3D/4D environmental data is critical to understanding and predicting environmental phenomena for relevant decision making. This research explores how to best utilize graphics process units (GPUs) and central processing units (CPUs) collaboratively to speed up a generic geovisualization process. Taking the visualization of dust storms as an example, we developed a systematic 3D/4D geovisualization framework including preprocessing, coordinate transformation interpolation, and rendering. To compare the potential speedup of using GPUs versus that of using CPUs, we have implemented visualization components based on both multi-core CPUs and many-core GPUs. We found that (1) multi-core CPUs and many-core GPUs can improve the efficiency of mathematical calculations and rendering using multithreading techniques; (2) given the same amount of data, when increasing the size of blocks of GPUs for coordinate transformation, the executing time of interpolation and rendering drops consistently after reaching a peak; (3) the best performances obtained by GPU-based implementations in all the three major processes, are usually faster than CPU-based implementations whereas the best performance of rendering with GPUs is very close to that with CPUs; and (4) as the GPU on-board memory limits the capabilities of processing large volume data, preprocessing data with CPUs is necessary when visualizing large volume data which exceed the on-board memory of GPUs. However, the efficiency may be significantly hampered by the relative high-latency of the data exchange between CPUs and GPUs. Therefore, visualization of median size 3D/4D environmental data using GPUs is a better solution than that of using CPUs.

Development of SD-OCT for Imaging the in vivo Human Tympanic Membrane

We report a novel extension of 840 nm wavelength- based spectral domain optical tomography to in vivo/real-time human middle ear diagnosis. The system was designed to access the middle ear region with a specifically dedicated handheld probe. The real-time displaying feature was mandatory for in vivo imaging human subject with the handheld probe, and the system could provide about 20 frames per second for 2048 pixels by 1000 A-scans without using any graphics process units under the Labview platform. The inner ear structure of a healthy male volunteer was imaged with the developed system with the axial and lateral resolutions of 15 µm and 30 µm, respectively. The application of the OCT technology to early diagnose otitis media(OM) is very promising and could be another extensive branch in the OCT field because it provides the depth resolved image including tympanic membrane (TM) and structures below TM whereas the conventional otoscope technique only gives asurface image of the TM.