DKFZ: A major step towards digital oncology
Methods of artificial intelligence (AI) such as deep learning using neural networks are gaining increasing importance in health research and personalized medicine. In oncology, new experimental and diagnostic methods such as genome sequencing and whole-body imaging by magnetic resonance imaging are also generating growing amounts of data.
Using AI to analyze these data makes it possible to identify unknown patterns and unexpected links. This enables scientists to make crucial contributions to cancer research and, thus, to assist physicians in diagnosis and treatment decisions. However, the data as well as the neural networks used nowadays have reached a complexity that is no longer manageable using conventional computer technology.
To tackle this challenge, the German Cancer Research Center (DKFZ) has now started operating a new infrastructure. It is based on two extremely powerful supercomputers which use graphics processing units for deep learning. For this purpose, each of these supercomputers has 16 times the memory of previous technologies and has a computing capacity corresponding to 600 classic processing units. This enables scientists to use mathematical models that are many times more complex than previous ones.
The DKFZ is the first German institution in the health sector to use this technology. Thus, it also runs the largest installation of this kind in the area of research and education throughout Germany.
A large number of departments and working groups at the DKFZ and the National Center for Tumor Diseases (NCT) will use the new server infrastructure for innovative research programs jointly with their collaboration partners. Goals are, for example, to develop the next generation of medical image and genome analyzing technologies as well as to advance computer-assisted surgery. Within the newly founded Helmholtz Information & Data Science School for Health (www.hidss4health.de), the installation also offers new possibilities in the education and training of young scientists.