Advances in gaming technology promise to improve medical imaging

A 65-year-old woman has presented to the ER with severe shortness of breath. She has been to the cath lab and needs intervention. She has a history of bypass to the right coronary artery, and the imaging results show a diffusing disease graft. Further observation indicates a tight 90% to 95% osteolesion and a 70% to 75% mid-lesion.

The wire filter case will be the first for the physician performing the procedure. He will manage the equipment and imaging controls while monitoring the responsive hemodynamics: ECG, arterial venous waveform, oxygen saturation, and blood pressure. If something goes wrong, he’ll have to respond, and chances are something will go wrong—the surgery is taking place in SimSuite, the simulation program from Medical Simulation Corp (MSC), Denver.

In this world, the patient gets extra lives, which permits inexperienced physicians to practice new procedures on mannequins rather than people. Not surprisingly, these technologies are often descended from those used in gaming, which has the funds to push the technical edge of research and development.

The advances occur in both hardware and software, such as graphics cards, processing units, and software development kits. “Gaming platforms are the kind we need but could never spend the same money developing, so we take advantage of what they have developed,” said John Lovett, product manager for the cardiology CT business unit of Toshiba America Medical Systems Inc, Tustin, Calif.

The technologies are applicable to not just simulation programs but also medical imaging. Toshiba has adopted gaming advances to improve the performance of imaging systems, such as the Aquilion CT. And companies like IBM, Armonk, NY, are teaming with medical institutions like the Mayo Clinic, Rochester, Minn, to develop new technologies that will increase imaging functionality. Medical imaging may not stay ahead of gaming, but medicine is using the consumer advances to raise its own game.

In the Game: Entertainment to the ER

Simulation is the most obvious application, in part because it most resembles a game. Medical games already exist, eg, Emergency Room: Code Red, which simulates the experience of an ER doctor, and 911 Paramedic, for a simulated EMT experience. “The gaming industry has made great advances in physics and graphics technologies in the past 5 years that are leveraged by the medical simulation companies,” said Dave E. Wilson, vice president of technology with MSC. Subsequently, medical simulation used for education and training purposes (as opposed to entertainment) also has improved tenfold, he suggests.

According to Wilson, this technology “sharing” has occurred since the 1970s but has increased in recent years. “The medical instrument industry is finally catching up to the fact that the gaming industry has a great deal to offer,” said Wilson. Gaming’s advances mean better performance: clearer graphics, faster speeds, and increased functionality.

“In a game, players see explosions or shootings in real time. Their actions are reflected on the screen immediately. In a medical simulation, you are not blowing things up, but you are placing a catheter and watching it move through the endovascular system—it’s the same thing, just a different context,” said Wilson.

Simulations can even incorporate mannequins (full or partial) to better mimic reality; programs can be used by new or experienced physicians who want to maintain their skills on little-used procedures.

Look Who’s Got Game

Users’ demands in the medical industry are similar to those of gamers. “They want very quick data delivery; complex, interactive, and realistic graphics; and interactivity, whether it’s with the adjacent computer through a LAN or another computer on another network across the world,” said Toshiba’s Lovett.

Toshiba writes its own software codes to run on the video cards, processors, and computers developed as a result of the gaming force in the industry. “We license the software development kits from gaming companies to develop proprietary technologies such as user interfaces, 3D imaging, and real-time capabilities,” said Lovett.

The new kits that emerge from gaming enable performance to go up and cost to come down. As memory speeds up, it gets less expensive. “We started with four gigabytes of memory, now offer eight gigabytes, and will probably increase to 16 gigabytes because it has gotten cheaper and faster,” said Lovett.

Resolution also improves. Lovett notes new systems can render a heart and arteries so that plaque is visible, even next to bone. Faster performance permits the imaging of beating hearts in two or three dimensions. And the technologies can be applied to any modality, including CT, MRI, and ultrasound. They must, however, also incorporate safety and security.

“We always deal first with patient safety and instrument efficiency and then security, which is not such a big issue in gaming,” said Lovett. Safety, efficiency, and security development testing do not get cheaper with advanced technology, but do become more effective, suggests Lovett. “Cost continues to go down as performance improves throughout the industry.”

This trend can be expected to continue, particularly as companies begin to look at gaming’s application to networking. “Doctors are not always at the hospital and will want to interact with the systems from anywhere using devices like the iPhone,” said Lovett. He expects advances from gaming technology to increase efficiency in networking speed and grid computing, suggesting new infrastructures will find their way into medical imaging applications.

Nontraditional Partnerships Put Imaging Ahead of the Game

IBM and the Mayo Clinic are collaborating on a program that significantly reduces the time needed for processing comparison data: the porting and optimization of Mayo Clinic’s Image Registration Application on the IBM BladeCenter QS20 “Cell Blade” produced image results 50 times faster than the application running on a traditional processor configuration.

The Cell or Cell Broadband Engine Architecture (Cell/BE or Cell BE) represents a significant advance in technology, featuring nine processing elements. The technology was developed jointly by IBM, Sony, New York City, and Toshiba in a venture known as STI. The technology’s best-known use is as the chip in the Sony PlayStation 3 (PS3). The Mayo Clinic and IBM have used the Cell’s parallel computer architecture and memory bandwidth to integrate different image types and view differences more accurately.

When monitoring treatment, physicians typically integrate information from four to six image types (eg, CT, MRI, etc). “With slices 2 mm to 4 mm wide, this quickly adds up to an overwhelming number of images,” said Bradley Erickson, MD, PhD, director of the Mayo Clinic’s radiology informatics lab.

Making comparison more complex is the fact that images from different time periods rarely reflect the same angle. “A patient almost never lies in the scanner in exactly the same way as before,” said Erickson, noting that even the tilt of the nose can confuse the images. The physician must then examine more closely whether the perceived difference in the image is due to the patient’s position or an actual difference in physiology or morphology.

Computer programs can perform this comparison automatically, adjusting for differing angles to align registration between the images. “A lot of the same types of computation used in games are also used in medical imaging manipulations,” said Erickson. However, because these calculations incorporate such large amounts of data, traditional processing has been slow.

The Cell BE architecture speeds the process. The Mayo Clinic/IBM research team ran 98 sets of images on the optimized registration system as well as a typical processor configuration and found large differences in processing time: the Cell BE architecture completed the registration for all sets of images in 516 seconds versus 7 hours on the typical setup.

The faster time to results means less anxious waiting for patients and more productive time to administer or alter treatment. Oncology is the primary area for this application, though Erickson suggests it is also appropriate for any chronic condition that must be monitored through imaging, such as multiple sclerosis.

Ideally, the algorithms will progress to the point where images from different time points can be condensed into one colorized image to show where pathology is improving or worsening. Erickson also would like to see it progress to commercialization.

Before this can happen, however, the organizations must work with regulatory agencies and manufacturers to assure the safety and efficacy of the program, particularly once integrated with other vendor systems. IBM and the Mayo Clinic have committed to combining resources, including programming and equipment, to the development of more tools for the imaging community. “The Cell processor at the heart of the PlayStation 3 is at the forefront of computing, and we aim to use the technology to further scientific and product development,” Erickson said. Let the games begin.

Renee DiIulio is a contributing writer for Medical Imaging. For more information, contact .

Better Imaging Starts with the Cell

Challenged by Sony and Toshiba to develop power-efficient and cost-effective high-performance processing for a wide range of applications, particularly high-performance gaming, IBM responded with the Cell Broadband Engine Architecture (aka Cell, Cell BE, Cell/BE).

Developed through a joint effort by the three companies, the Cell is a heterogeneous chip multiprocessor comprised of an IBM 64-bit Power Architecture core, augmented with eight specialized co-processors based on a novel single-instruction multiple-data (SIMD) architecture called Synergistic Processor Unit (SPU), which can handle the data-intensive processing required for scientific applications.

The architecture is scalable and modular, and potential applications include not only gaming and medical imaging, but also HDTV sets, home servers, and supercomputers.

Cell statistics:

  • Observed clock speed: > 4 GHz
  • Peak performance (single precision): > 256 Gflops
  • Peak performance (double precision): > 26 Gflops
  • Local storage size per SPU: 256KB
  • Area: 221 sq mm
  • Technology: 90 nm SOI
  • Total number of transistors: 234M

For more information, see The Cell Architecture: Innovation Matters. IBM. Available at domino.research.ibm.com/comm/research.nsf/pages/r.arch.innovation.html.

—Ann H. Carlson