Resources

Center for Intelligent Imaging Computational Infrastructure Resources

 

Computational Core: Infrastructure and Resources

The UCSF Center for Intelligent Imaging computational core maintains responsibility for developing and managing state-of-the-art computational hardware and software to enable cutting-edge biomedical imaging research and clinical translation. The Center aims to provide internal and external researchers and clinicians with the most advanced CPU and GPU computing archiecture, large-scale data storage and interoperability tools, and software stack for both traditional image analysis and state-of-the-art machine learning techniques.

The computational core focuses on software development and implementation, core competence in data science, and CPU and GPU high performance computing. Together with UCSF's state-of-the-art high performance computing centers and a uniquely extensive clinical data record, the core provides the resources to support advanced quantitative imaging research and to develop applications that can be readily integrated into the clinical radiology workflow to directly improve patient care. Current areas of research include focus on automation and quantitation of decision support, image acquisition and reconstruction, quantitative analysis, workflow efficiency improvement, and outcomes prediction.

Computational Infrastructure

The ci2 computational and infrastructure cores leverage CI, data and data science platforms developed within the department of radiology and across campus to support the center’s work including:

Radiology Scientific Computing Services 

SCS supports the computational core via recharge subscription to Radiology and Biomedical Imaging research groups.

SCS

High Performance Computing

  • The UCSF Department of Radiology and Biomedical Imaging maintains its own high-performance computing environment as well, utilizing Sun's Grid Engine (SGE) scheduler to support batch CPU and GPU job submission. That grid comprises approximately 1000 CPU cores and 46 GPUs.
  • Nvidia's DGX-2, the world's first petaFLOPS system combining 16 interconnected GPUs is available to accelerate machine learning. 
  • UCSF's shared high-performance computing cluster, also known as Wynton, comprises approximately 4000 CPU cores and 116 GPUs.

    Wynton

Software

  • Software provided by external collaborators in both academic and commercial sector combines with in-house algorithms developed in the course of research.
  • Clinical validation platforms for prospective evaluation of new quantitative software
  • Image annotation: MD.ai, NVIDIA AIAA

Data and Data Science Platforms

Information Commons is a fast and easily searchable and accessible repository of all UCSF clinical data and models, and related basic science & population data, that enables UCSF research and discovery of new health insights underlying precision medicine, to improve patient and health care.

Information Commons

Clinical Radiological Imaging Data

AIR

Center for Intelligent Imaging Data Resources