News

University of Groningen

 
 

The Zernikeborg building, where the University of Groningen's Centre for High Performance Computing and Visualisation is housed

The Center for High Performance Computing and Visualisation (www.rug.nl/cit/hpcv) is part of the University of Groningen and offers scientists state-of-the-art facilities for virtual reality, visualisation and supercomputing. It houses the fastest computer in the Netherlands, a 12288 core BlueGene/L. However, neuroscience researcher Dr Hans van Hateren has found that PC-based graphics processing units (GPUs) from NVIDIA provide better performance for his work to simulate neuronal responses in the human eye.



challenge

By simulating retinal circuits, the components of the eye which actively process visual information and pass it to the brain, Dr van Hateren’s research aims to provide new insights into the functioning of the human eye. A greater understanding of retinal function is an important step in the development of artificial models of the human visual system, which would have a wide range of important medical, scientific and industrial applications. These include the design of retinal implants to help improve the vision of people with sight problems, and the development of visual systems for robots.

Image processing by the human retina is massively parallel – in other words, there are many pieces of information being processed simultaneously. Large numbers of cones (light sensitive cells which form part of the retina) sample the image projected onto the retina into three colour channels and adapt to prevailing light levels. Other neurons in the retina provide spatial interactions that sharpen edges and enhance color contrast in the image.

Simulating these processes is computationally very demanding. It is well suited for parallel computer architectures, but requires fast and efficient communication between different processors because of the extensive spatial interactions in the retina.

solution

A drawing of a section through the human eye, with a schematic enlargement of the retina. NVIDIA's technology is used in research to simulate the complex processes which take place in these structures. Image reproduced courtesy of www.webvision.med.utah.edu.

NVIDIA® professional graphics solutions provide an excellent platform for performing these computations. The GPU’s inherently parallel design means large numbers of processors act simultaneously, making it ideally suited to handle computationally intensive data. Dr van Hateren explains: “With NVIDIA’s graphics technology, it’s possible to use fast algorithms that closely mimic the complex physical and chemical processes occurring in the human retina. The excellent performance growth curve of successive generations of cards, the support for Linux, and the recent addition of the CUDA™ software environment enables continuing cost-effective, up-to-date high performance computing.”




impact

Dr van Hateren’s efforts to better understand the human visual system have been advanced by NVIDIA technology. “The high performance of NVIDIA’s solutions means we’re able to create simulations which very accurately match the way retinal neurons behave. Not only has NVIDIA technology been instrumental in the development of this retinal model, its high performance allows us to mimic some of the complex processes involved in human sight. For example, we are able to render and compress High Dynamic Range video in a very similar way to that used by the eye to process images. The results of these simulations are providing new insights into how we see.”



 
 
FacebookTwitterLinkedInGoogle+Pinterest