“OK, let’s do it,” says a tall man in pale green surgical scrubs. “Lights please.” As the lights fade, Stefan Brew stands silhouetted by the glow from a bank of X-ray monitors. In front of him a child’s hair peeps from under a blue drape that gently rises and falls as the young patient breathes.
All eyes focus on a pair of X-ray monitors showing the tree-like arrangement of arteries and veins in the boy’s brain. Suddenly, a handful of clumpy vessels turn white as a glue-like substance streams into them and seals them off. A few minutes pass before Brew announces to his team: “Done, thanks. It looks all right. He’s in pretty good shape.”
James, the 8-year-old patient, suffered a stroke in February and required emergency surgery to drain a brain haemorrhage. Then last month, at London’s Great Ormond Street Hospital for Children, Brew’s team found the cause: James had a dangerous arteriovenous malformation (AVM), an abnormal cluster of potentially leaky connections between veins and arteries. During the procedure described above – called an embolisation – Brew sealed off the AVM by injecting a hard-setting plastic liquid through a catheter. The liquid is either an agent called onyx or else the superglue chemical cyanoacrylate, both of which set rapidly on contact with blood and block the abnormal vessels.
Advertisement
It sounds simple, but the brain’s vasculature is complex and varies between patients. Doctors admit that working out which vessels to seal off is often a matter of trial and error, and around 1 in 10 patients suffer haemorrhages as a result of embolisation procedures.
“At the moment we do AVMs more or less blind,” says Brew, an interventional neuroradiologist at the National Hospital for Neurology and Neurosurgery in London. “We eliminate as much of it as we can, but it’s basically a blunderbuss approach.”
Now that promises to change. A new imaging system aims to give surgeons a detailed, near real-time picture of the vasculature – and its abnormalities – and so remove much of the guesswork.
Called “Grid-enabled neurosurgical imaging simulation”, or Genius, the system fires off brain scans to a network of supercomputers. These then use their collective processing might to generate an accurate 3D model of the unique blood flow patterns in a patient’s brain, showing the surgeon visual representations of critical parameters such as blood pressure and flow rate.
These models, created just prior to each operation and updated in near real-time, will also let doctors test the effect of a particular intervention before they actually do it. Within a year the system is expected to be part of the world’s first brain procedure guided by a real-time computer simulation.
The imaging side of Genius is being made possible by a brain scanning technique called rotational 3D angiography. This builds up a 3D image of the vasculature from large numbers of 2D X-ray slices shot from different angles around a 180-degree arc. The resulting model can be viewed from different angles, making it easier to spot vascular problems.
However, the shape of the blood vessels does not reveal the key pressures and flow velocities that may indicate where ruptures are likely. So Marco Mazzeo and Peter Coveney of University College London have written software that calculates these critical parameters, as well as stresses on artery and vein walls, based on a handful of pressure measurements made by the surgeon using a special catheter. That data can then be turned into colour-coded images so surgeons can see stress points.
Surgeons working on a patient also need such images to be bang up to date – not an old snapshot that may not reflect changes due to disease or physical injury, say. That means processing vast amounts of data to constantly update the model – around a trillion calculations per second.
To do this, Genius will access 20 supercomputers across the US’s TeraGrid and the UK’s National Grid Service, which offer a dedicated processing infrastructure for scientists. Special software will let surgeons make advance reservations for this processing time, or give them immediate access in emergencies.
“These procedures involve life and death decision-making,” says Coveney. “The goal is to enhance the ability of clinicians to make these decisions through information technology and high-performance computing.”
According to its developers, who also include teams at the Universities of Manchester and Edinburgh, UK, Genius will help surgeons deal with the three most common problems affecting neurovasculature: aneurysms, atherosclerosis and AVMs. Aneurysms are balloon-like swellings of vessels in the brain. If they rupture, blood pours into the fluid around the brain, depriving cells and tissue of oxygen and nutrients, and potentially triggering strokes.
Computing the risk
Experts believe that more than one in 100 people have a brain aneurysm. Most suffer no ill effects, but of those whose aneuryms rupture, around half will die as a result. Because doctors can’t easily tell whether an aneurysm is likely to rupture, this leaves them in an unenviable position: should they intervene or not? “If you tried to treat the millions of aneurysms around the country, you would do more harm than good,” says Brew. Supercomputer modelling will tell surgeons whether an aneurysm is likely to bleed, and if there is more than one aneurysm, which poses most risk.
Atherosclerosis, the hardening and narrowing of arteries, also causes strokes. Normal procedure now is to insert a stent to expand the vessel and restore blood flow, but in many cases the narrowing occurs at more than one location. This and the pulsed flow and variable viscosity of blood make predicting the outcomes of such procedures highly complex. Here again, Genius will help doctors customise their interventions.
Brew believes the system’s greatest potential is in the treatment of AVMs. “Modelling could allow us to identify the problem and deal with it in a safer, more focused way,” he says.
Nor is Genius alone. It’s part of a wider effort to model organs or body systems under the umbrella of the Virtual Physiological Human (VPH) project. Other VPH efforts include a simulation of the musculoskeletal system by Marco Viceconti of the Rizzoli Orthopaedic Institute in Bologna, Italy; the Giome project to model the gut, coordinated by Danish researchers; the Renal Physiome Project to develop a virtual kidney, which is being spearheaded by French scientists; and an attempt in New Zealand to model the workings of the heart.
Last October the European Union allocated ¬72 million to VPH projects to help turn these efforts into clinical applications and establish a common framework that will one day allow them to be stitched together into a virtual human. The full price tag of the VPH vision has been estimated at ¬500 million.
Brew sees this approach as a major way to reduce risk and take fewer gambles, likening it to playing an intelligence-based game like chess instead of more chance-based poker.
“In chess the greater application of logic will always win out,” he says. “While poker is also a game of skill, there is also a significant element of uncertainty and chance. Modelling has the potential to make these procedures more like chess and less like poker.”
The Human Brain – With one hundred billion nerve cells, the complexity is mind-boggling. Learn more in our cutting edge special report.
