Challenge: As a leading academic medical center and research institution, the UCSF CDHI is improving clinical care through innovation in artificial intelligence (AI). To advance that work, university researchers require tools and cost-effective infrastructure to create new algorithms, train AI models, and run a variety of workloads. For the OA use case, UCSF researchers needed to develop and ...train a deep learning model that could examine sophisticated 3D medical images and identify those indicating torn knee cartilage.
Solution: The UCSF CDHI used BigDL on Apache Spark* to develop algorithms and train models, and worked with Intel, Dell, and Cloudera to deploy a data analytics cluster based on the Intel® Xeon® processor Scalable family. Taking advantage of Intel data center technologies, UCSF broadens its researchers’ ability to develop algorithms and models as well as to import and optimize models from other AI frameworks.
Results: The OA team is meeting its first-phase accuracy goals, demonstrating progress on a clinically significant problem with long-term potential to improve diagnosis, treatment planning, and clinician productivity. Intel’s high-performance platform and tools are helping UCSF researchers speed time-to-results and make high-performance, AI capabilities broadly available to researchers.