Focus points for the Computer Graphics and Scientific Computing group

Graphics cards deform heart tissue

Researchers are using the large amount of processing power present in a graphics card to create a surgical simulator for doctors.

How do you create a computer visualisation of a congenitally malformed heart and a simulation of its repair of such good quality that the surgeon is guaranteed not to receive any unpleasant surprises during the operation?

The answer is by using powerful graphics cards of the type that hardcore gamers have in their computers.

And the answer comes from the research group Medical Visualisation and Simulation, which is part of Computer Graphics & Scientific Computing at the Department of Computer Science at Aarhus University.
The newest graphics cards can both create advanced visualisations and physical simulations – the latter most often many times faster than the computer’s CPU, which researchers around the world have been using previously.
In this way it is possible to create real time 3D graphics based on the images from e.g. CT, PET or ultrasound scanners.

Elastic tissue
Tissue is elastic and changes form when it is pushed or pulled. This is also the case for the 3D renderings – that is when you are using digital scalpels, forceps and tongs.
The surgeon can thereby make surgical incisions, “pull” heart tissue to the side and look into the patient’s heart chambers on his computer and in this way carefully plan the correct operation on individualized heart models.
The group has just conducted a study at a children’s hospital in London to investigate whether the patient-specific 3D models can be created in a reasonable period of time and are sufficiently accurate.

”We tested it on 40 patients and it was demonstrated that we were able to create a sufficiently accurate model of a patient’s heart in an average of an hour. The surgeons can therefore have faith in the fact that they will also be able to see what they see in the 3D model when they open the patient's heart”, says Assoc. Prof. Thomas Sangild Sørensen, who is a computer scientist and has a PhD in medicine.

“Many different groups in the world are working with surgical simulation and one of the greatest challenges is to perform the simulation in real time while at the same time maintaining the level of detail sufficiently high. We have found a solution to the tissue deformation problem, which can be resolved by using parallel calculations on the graphics cards; they have a large number of processors that can each deal with their tasks simultaneously and therefore everything happens e.g. 50 times faster than using just the CPU – you see, the new gaming graphics cards have several hundred processors”.

 “We were among the first to use graphics cards as parallel processors for surgical simulation”, adds Thomas Sangild Sørensen, who together with three PhD students, Karsten Østergaard Noe, Allan Rasmusson and Lau Brix, currently make up the research group Medical Visualisation and Simulation. A former PhD student, Jesper Mosegaard, who is now employed at the Alexandra Institute, participated in the development of the cardiac surgical simulator.

Cycling in the scanner
Most recently the group has worked intensively on advanced real time reconstruction of real time 2D image data from MRI scanners.
Currently the best you can achieve from an MRI scan is a couple of images per second of poor resolution. Alternatively the patient holds his or her breath for ten seconds in order to achieve a high resolution image. If we wish to record and reconstruct images with both good spatial and temporal resolution, complicated mathematics and an effective implementation are required in order to reconstruct the images as fast as data is recorded.

“Our vision is to be able to create 20 images per second of high resolution. We can already do this now in the laboratory using our algorithms and graphics cards. This will open up a wealth of new clinical opportunities, especially since we will be able to see sharp images of moving objects. For example, you could “cycle” inside the scanner and the scanner will be able to show how your physiology changes”, says Thomas Sangild Sørensen.

The group is collaborating with computer scientists, engineers, physicists and doctors at hospitals in Aarhus, Bordeaux and London and expects that the system will be able to be run on all scanners in the future.