Siemens Healthineers with Intel® Xeon® processors

4th Gen Intel Xeon processor allows Siemens Healthineers to boost medical image processing, and meet sustainability efforts.



I'm Sasa Grbic. I'm a Senior Director at Siemens Healthineers, and I'm heading the Whole Body and Oncology Research Group. At Siemens Healthineers, we are committed to helping health care professionals deliver quality care and improve patient experience and outcomes by transforming care delivery and precision medicine with artificial intelligence.

Contouring organs at risk is an essential step during the planning phase for radiation therapy. It consists of radiation therapy professionals manually contouring tens of organs at risk on a computed tomography data set. This process is tedious and time-consuming, and the resulting contours often lack consistency because contours can differ from user to user.

Supporting radiation therapy professionals with AI based auto-contouring technology can increase workflow efficiency, improve consistency, and help free up staff to focus on additional value adding work. High performing auto-contouring technologies for organ at risk require convolutional deep neural networks with millions of parameters. As complexity of these neural networks has increased, on today's third generation Intel Xeon Scalable Processor architecture, the runtime can be substantial, thus, limiting the applicability of these algorithms in interactive applications. The newly introduced fourth generation Intel Xeon Scalable Processor is well-positioned to address this challenge.

I am Bogdan Georgescu, and I'm a Principal Key Expert Scientist in AI-based machine learning analysis of images at Siemens Healthineers. As part of our research, we at Siemens Healthineers benchmarked the AI inference time for auto-contouring algorithms on the new fourth generation Intel Xeon Scalable Processor, and achieved a 35 times speedup relative to running it on a third generation Intel Xeon Scalable Processor. The AI inference for contouring a typical abdominal scan with nine structures takes only 200 milliseconds.

This AI acceleration is made possible on a fourth Gen Intel Xeon Scalable Processor with microarchitecture advancements and integrating accelerators like advanced matrix extensions. These CPU features are leveraged by the Intel OpenVINO toolkit to speed up the execution of AI models. The hardware and software acceleration available within fourth generation Intel Xeon Scalable Processor allows Siemen's Healthineers to further lower system cost and complexity as well as energy consumption by up to 20%.

Breakthroughs in health care AI requires years of research using large data collections, as well as powerful and reliable supercomputing framework. Intel provides an extended supply availability of up to 10 years for the fourth generation Intel Xeon Scalable Processor, which enables Siemen's Healthineers to maximize value of investments by ensuring uninterrupted infrastructure for continuous research and innovation.