ACADEMIA
Universite de Montreal Researchers Simulate World's Largest Heart Model
1.2TB SGI Altix Shared Memory System Runs 'Heart Model of the Future' Today, Accelerating Researchers' Time To Insight: In the quest to discover how the mechanisms of disease work, researchers at the Universite de Montreal (UdM) have run the largest mathematical simulation of a heart ever assembled -- a 2 billion element model -- on a high-performance computing system from SGI. The new UdM model is up to 1,000 times more detailed than previous models, enabling new scientific discoveries that would never be possible via observation alone. Until recently, the largest heart models in the world had at most a few million elements. Over the last nine months, Dr. Mark Potse and Dr Alain Vinet, both affiliated with the Research Center of Sacre-Coeur Hospital and the Biomedical Engineering department at UdM, began running 100 to 120 million-point models as part of their heart disease research on an SGI Altix 4700 system, believed to be the largest shared memory computing system in Canada. They regularly use 60 of the 768 Intel Itanium 2 processors running on the SGI Altix which, as part of the Quebec Network for High-Performance Computing (RQCHP), is shared by many researchers from across Canada. 
In late October, Potse and Vinet had the opportunity to run their custom electrocardiography (ECG) code to solve the largest, most detailed heart model ever, using the entire SGI Altix system and 1.2TB of shared memory. Originally written by them on an older SGI system and ported to the SGI Altix system's Linux environment in 2003, the ECG code made the leap from 120 million points to 2 billion with ease.
"We have been using the model code for research and not really developing it further, but after the success of the trial I am now thinking about improving the model, making it much larger and much more detailed, and attacking other diseases that we couldn't handle before," said Dr. Potse. "It's a very complicated model and it's much, much easier to write parallel programs on a shared memory machine. The Altix delivers fast processing performance for our application needs; not only because of the shared memory, but also due to the very high bandwidth interconnect. That's good for the kind of mathematical equations I'm solving and saves me a lot of time."
Potse simulated 5 milliseconds of activation in a tissue block that included some properties of a real heart, such as fiber running in different directions. The simulation solved a system of 2 billion equations a dozen times. The test took two hours, which Potse describes as short for achieving the desired results. A full heartbeat, he says, would take two weeks, and they cannot claim use of the entire machine for that length of time right now.
"This was a test to see if the simulation works and to determine that, if we have a much bigger machine, our software will be able to run more efficiently," added Potse. "This capability is really for the future when we can use this size of machine on a regular basis, but with the Altix system we have made the heart model of the future today."
With heart disease one of the leading causes of death in the Western world, discovering the electrical "triggers" of the various kinds of heart disease could lead to earlier diagnosis and new treatment breakthroughs. For example, there are inheritable diseases such as "Brugada syndrome" and "Long-QT syndrome" which can cause sudden death in young, otherwise healthy people, and there are typical changes in the ECG that allow doctors to diagnose these diseases. In order to understand what the mechanisms of the particular disease are, the heart must be modeled with enormous detail. Once disease mechanisms are fully understood, scientists will be able to devise the best drug or the best cure -- surgical or other remedies -- and doctors will be able to diagnose much more precisely.
Without the use of computer models it can be hard to track the effects of a heart disease on the ECG. For instance, many patients with Brugada syndrome have an inheritable disorder on the level of ion channels -- large molecules in the cell membrane that help activate the heart in order to make it beat. It is tempting to believe that these ion-channel disorders cause the typical ECG and the risk of sudden cardiac death that are linked with Brugada syndrome. However, computer simulations have shown that this is not the case. Additional factors are needed. Such a discovery, which disproves a proposed explanation, could only be made with a detailed computer model of the heart, not by observation alone. Moreover, the level of detail required for this application demands such large computational resources that it could be achieved only very recently.
"For 25 years SGI has fueled biomedical innovations, accelerating scientific research by reducing time to insight in projects as varied as the mechanics of HIV protease, genomic correlation for cancer research, and 3D simulations of surgeries," said Michael Brown, Director of Sciences Markets, SGI. "The SGI Altix system now bolsters the future of heart disease research by proving that expanded calculations of 2 billion elements are not only possible, but will become the norm."
In late October, Potse and Vinet had the opportunity to run their custom electrocardiography (ECG) code to solve the largest, most detailed heart model ever, using the entire SGI Altix system and 1.2TB of shared memory. Originally written by them on an older SGI system and ported to the SGI Altix system's Linux environment in 2003, the ECG code made the leap from 120 million points to 2 billion with ease.
"We have been using the model code for research and not really developing it further, but after the success of the trial I am now thinking about improving the model, making it much larger and much more detailed, and attacking other diseases that we couldn't handle before," said Dr. Potse. "It's a very complicated model and it's much, much easier to write parallel programs on a shared memory machine. The Altix delivers fast processing performance for our application needs; not only because of the shared memory, but also due to the very high bandwidth interconnect. That's good for the kind of mathematical equations I'm solving and saves me a lot of time."
Potse simulated 5 milliseconds of activation in a tissue block that included some properties of a real heart, such as fiber running in different directions. The simulation solved a system of 2 billion equations a dozen times. The test took two hours, which Potse describes as short for achieving the desired results. A full heartbeat, he says, would take two weeks, and they cannot claim use of the entire machine for that length of time right now.
"This was a test to see if the simulation works and to determine that, if we have a much bigger machine, our software will be able to run more efficiently," added Potse. "This capability is really for the future when we can use this size of machine on a regular basis, but with the Altix system we have made the heart model of the future today."
With heart disease one of the leading causes of death in the Western world, discovering the electrical "triggers" of the various kinds of heart disease could lead to earlier diagnosis and new treatment breakthroughs. For example, there are inheritable diseases such as "Brugada syndrome" and "Long-QT syndrome" which can cause sudden death in young, otherwise healthy people, and there are typical changes in the ECG that allow doctors to diagnose these diseases. In order to understand what the mechanisms of the particular disease are, the heart must be modeled with enormous detail. Once disease mechanisms are fully understood, scientists will be able to devise the best drug or the best cure -- surgical or other remedies -- and doctors will be able to diagnose much more precisely.
Without the use of computer models it can be hard to track the effects of a heart disease on the ECG. For instance, many patients with Brugada syndrome have an inheritable disorder on the level of ion channels -- large molecules in the cell membrane that help activate the heart in order to make it beat. It is tempting to believe that these ion-channel disorders cause the typical ECG and the risk of sudden cardiac death that are linked with Brugada syndrome. However, computer simulations have shown that this is not the case. Additional factors are needed. Such a discovery, which disproves a proposed explanation, could only be made with a detailed computer model of the heart, not by observation alone. Moreover, the level of detail required for this application demands such large computational resources that it could be achieved only very recently.
"For 25 years SGI has fueled biomedical innovations, accelerating scientific research by reducing time to insight in projects as varied as the mechanics of HIV protease, genomic correlation for cancer research, and 3D simulations of surgeries," said Michael Brown, Director of Sciences Markets, SGI. "The SGI Altix system now bolsters the future of heart disease research by proving that expanded calculations of 2 billion elements are not only possible, but will become the norm." 