GPU-Compute Server

• Subject Area: Medicine ; Computer Science
• Term: Funded in 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 530008379

We are applying for funds to procure the necessary hardware to set up an infrastructure for artificial intelligence (AI) in integrated image-based medical research. The infrastructure to be set up consists of a high-performance and efficient GPU-based compute system, which is required for the development of validation of AI methods in image-based medical sciences (radiology and pathology). An associated storage system is connected to it, that can provide the training data at high speed. This AI system is essential for data analysis in radiology and pathology, as very large data sets often have to be processed and analyzed (big data). We currently produce around 600 TB of data (pathology, radiology, molecular pathology) annually. The manual (clinical) evaluation of this multidimensional data is becoming increasingly complex and time-consuming. At the same time, this image data potentially has a very high information content for the respective disease subtypes and the resulting individualized, highly specific therapy options. The automated, AI-supported analysis of this data is already showing great success in helping doctors to make more accurate diagnoses and work more efficiently. Furthermore, new biomarkers can be found with AI, which enable the early diagnosis of diseases such as cancer or improve the individual therapy and care of patients through more precise subtyping. With the requested infrastructure, we want to deepen these research areas, investigate deeper patterns in the medical image data and thus contribute to better patient care and a better understanding of the mechanisms of pathologies. So far, there is no existing AI infrastructure for this research area.

• DFG Programme: Major Research Instrumentation
• Major Instrumentation: GPU-Computingserver
• Instrumentation Group: 7000 Datenverarbeitungsanlagen, zentrale Rechenanlagen
• Applicant Institution: Technische Universität München (TUM)
• Leader: Professor Dr. Peter Schüffler