By Stuart G. Hall
The impact of advances in gaming industry technology could herald a major shake-up in medical imaging, with patients, clinicians, and manufacturers all emerging as winners.
The decades-long dependence on the IT industry for improvements in processor power has been overturned by gaming processing innovations, with millions of gaming customers providing the rich economies of scale needed to kick start this mini-revolution. The economics of the gaming industry do two things; firstly it makes it economically feasible so it gets built at all. And secondly those volumes drive the cost of it down - a conclusion backed by the 2006 Norden report (see below). Which means for one thing that clinically-valuable applications that previously weren’t economically feasible are now viable.
The Cell Accelerator Board (CAB) is a PCI Express® accelerator card based on the Cell Broadband Engine(BE) processor.
At the epicenter of this change is the Cell Broadband Engine™ (BE) processor designed for the Sony PlayStation 3. Technically by the nature of the technical design of the chip for the PS3 it is very good at many of the basket of physics calculations, and medical imaging relies on the algorithms that also use a similar basket of physics calculations.
The subject of computing power is a hot topic at the moment with the forthcoming June SNM Annual Meeting in Washington, D.C. which also focuses on the explosive growth in the power of digital computers used for molecular imaging and nuclear medicine. Hard on its heels is the late July ‘Workshop on Solving Computational Challenges in Medical Imaging,’ where the focus is on how next-generation technology, currently featured in computer entertainment and video-processing platforms, is helping drive advancements in medical imaging.
“Now we have the alternative to take from the gaming industry. And it turns out that as they’re getting more sophisticated in what they’re trying to accomplish, their algorithms are much more similar to what we are trying to enable than the algorithms the old IT chips were focused on,” says Mercury Computer Systems’ Joel Radford in an interview with Medicexchange. Over the past Mercury had been dependent on chips developed first and foremost for the IT market: “What’s happening with gaming is that the many millions of units in gaming are creating a whole set of competing economics to drive differently architected processors.”
In addition Mercury’s own volumes are not high enough to drive such innovation, which means they have to leverage such gains from other industries in use in medical imaging: “It’s still early but it’s a huge paradigm shift for what is possible for our customers to do with their applications because of the availability of these new processors.”
Mercury Computer Systems, which is headquartered at Chelmsford, Massachusetts, has been working with the Institute of Medical Physics (IMP) in Germany on the joint development and commercialization of medical imaging technology deployed on the Cell processor. In parallel Mercury is engaged with some of the household names in the medical imaging industry using the Cell, and it is under evaluation for CT for digital X-Ray and other modalities.
“We started working with the Cell technology at the OEMs 12 to 18 months ago with the result that they’re now getting to the stage where they are designing it into the next generation of equipment, so it will still be some time yet before it’s out there in a clinical setting,” reports Radford. He added that today Mercury works with its customers on a number of acceleration choices such as the Cell, but today, for certain medical imaging algorithms those choices are not as attractive due to worries about the power footprint, the cost, the space constraints, and ease of porting applications across effectively.
“We’ve found on a chip-to-chip comparison basis performance gains of 100 times improvement in some cases. If you take that improvement in the context of what a medical OEM is attempting to accomplish it means they can put a ton of computer power in a small space for a small cost. It’s about price-performance.” But the real difference in clinical terms is real-time evaluation feedback Radford says: “Even before we worked with our colleagues in Germany we took a CT reconstruction algorithm and ported it to Cell almost two years ago now and achieved the 100 times increase, which means the 3D image literally pops up on screen in a few seconds. The competing chip is something like five to eight minutes.”
Related Reading
Medical Imaging and Computer games May 2006.
