Interactive modelling and visualisation to become reality with SCI-run solutionMannheim 08 jun 2000 During the afternoon session of the Supercomputer 2000 Conference, Dr. Chris Johnson, head of the Scientific Computing and Imaging (SCI) Institute at the University of Utah, addressed the topic of ASCI-machines used for modelling and visualisation in cardiology and neuro-surgery, for adaptive finite elements applied to atmospheric research, and for fire simulation. Five university centres in the USA are involved in the ASCI project. The architecture used in this scientific research constitutes the ASCI Blue Mountain, hosted at Los Alamos National Laboratories.
The research group for Computer Technology, led by Dr. Johnson, has created a 3D torso model from magnetic resonance imaging (MRI) scans. The team performed a so-called computational pipeline, meaning the automatic execution of segmentation and tissue classification for geometric modelling, large volume shaping with mesh generation, and application of boundary conditions. Today, the whole operation is still realised in batch mode, starting from modelling over simulation to visualisation and repeating the process over again until the desired results are achieved. Dr. Johnson has a dream of making this complex procedure user interactive so that the researcher or scientist can enter the process anytime to introduce corrections. To that purpose, the team is developing the SCI-run solution to generate a parallel structure.
The medical areas in which the system is bound to pay large services are cardiology and neuro-surgery. The problem of defibrillation device design can be tackled with 4D visualisation. When a person suddenly has a cardiac arrest, the heart is shocked with defibrillation equipment in order to regenerate it. Dr. Johnson compared the heart attack with a Windows computer break down. One has to restart the system. Only, in contrast with a computer, the patient has no restart buttons so defibrillation is required. The newest method of regeneration involves the implantation of electrodes into patients at risk. Advanced medical image modelling and visualisation shows the surgeon the exact location where to place the electrodes. An Origin 2000 system calculates the vector fields and allows a partial re-creation of the mesh without a need to restore the complete mesh.
Neuro-surgical interventions in turn usually are very time-critical. Dr. Johnson's team actively collaborates with Harvard and Brigham Women's Hospital in experiments where the patient is submitted to surgery while staying in the MRI scanner, enabling the surgeon to simultaneously consult the patient's medical images. Here, the final objective is to achieve real time simulation to safely guide the surgeon in the midst of the most complex interventions. The SCI team already applied inverse localisation to detect the neural activity in a patient's brain.
A non-medical application of the SCI-run system regards the time-dependent adaptation of finite elements for atmosphere research. To this goal, the Samourai system has been integrated into the SCI-run solution, allowing to perform parallel computing. In the Uintah project on the other hand, fire simulation and combustion research form the subject of advanced distributed computing. Parallel resources are used to set up complex simulations of fluid models. The smallest simulation requires no less than 10 billion degrees of freedom. "Speaking of a real challenge, there you have one", as Dr. Johnson pointed out.
The Utah team already succeeded in achieving promising and successful results in interactive visualisation for medical applications, scientific computing, and geoscience. The current computing performance amounts to 38 MFlops per processor. The SCI-run system allows to apply a view-dependent approach for large data sets and equally integrates haptics feedback. The requirements consist of OpenGL, Tcl, and p-threads. At this point, the system is not yet available for actual use but Dr. Johnson confirmed that an open source will soon see the light of day. The principal task for the team is to finish the documentation.
The real asset of the SCI-run system lies in its pipeline complexity, its potential for interactive simulation, and the possibility for visualisation everywhere instead of the burdensome procedure of post-processing. More information on the activities of Dr. Johnson and his research group can be found at the Web site of the Scientific Computing and Imaging Institute.
Leslie Versweyveld |
 |