Siemens Healthineers' innovative technology uses AI for real-time cardiac MRI diagnosis

Intel and Siemens Healthineers* are collaborating on a breakthrough AI-based cardiac MRI (magnetic resonance imaging) segmentation and analysis model that has the potential to provide real-time cardiovascular disease diagnosis. Intel and Siemens Healthineers used 2nd Generation Intel® Xeon® Scalable processors for AI inference to provide real-time magnetic resonance imaging (MRI) inference results to technologists, cardiologists and radiologists.

“Siemens Healthineers and Intel share a common goal to use AI to improve healthcare,” said David Ryan, general manager of the Life Sciences and Health Division of Intel’s Internet of Things Group. “By deploying 2nd Generation Intel Xeon Scalable processors with Intel® Deep Learning Boost and the Intel® Distribution of OpenVINO™ toolkit at the edge, data will be analyzed the moment it is acquired, enabling real-time cardiac MRI applications.”

Why this technology is important:

Cardiovascular disease causes one in three deaths in the United States—34 deaths per minute, or 18 million deaths per year.1 Cardiac MRI has become the gold standard for assessing cardiac function, ventricular volumes, and myocardial tissue.1

Cardiologists typically use manual or semi-automated tools to extract quantitative measurements from cardiac magnetic resonance imaging (CMR), but this step is time-consuming and error-prone, and is subject to subjectivity in interpreting the images.

“Based on Intel Xeon Scalable processors, we can now develop multiple real-time and mission-critical medical imaging use cases, such as cardiac MRI, without the added cost and complexity of additional hardware accelerators,” said Dorin Comaniciu, senior vice president of Siemens Healthineers.

The AI-based heart model will save cardiologists more time, eliminating the need for them to manually segment the ventricles, myocardium, and cardiac blood pool. When the scanner generates image slices, AI-based image segmentation is performed immediately at the edge, allowing the computing system deployed at the edge to capture the generated data in real time - this brings advantages such as low latency and high throughput to AI reasoning, allowing medical institutions to safely provide diagnostic services to more patients every day.

What are the benefits of this technology:

The life sciences and health industry is undergoing a digital transformation in healthcare, using artificial intelligence to speed up clinical workflows, improve accuracy and diagnostic levels, and provide greater support for medical research while reducing hospital costs. Artificial intelligence can quickly provide visualization of anatomical systems and identify abnormal conditions, which helps clinicians further focus on patient care.

About technology:

Currently, most systems deployed by Siemens Healthineers use Intel® processors, which enables Siemens Healthineers to leverage existing CPU-based infrastructure to run AI inference workloads. Siemens Healthineers and Intel use the Intel® Distribution of OpenVINO™ toolkit to optimize, quantize, and execute models. The final demonstration results show that the speed has been increased by more than 5 times with almost no loss in accuracy2.

Intel® Deep Learning Boost is a new set of embedded processor technologies that accelerate the implementation of deep learning use cases. It extends the new Vector Neural Network Instructions (VNNI) in the Intel® AVX-512 instruction set, which is built into the second generation Intel Xeon Scalable processors. In the past, tasks such as convolutions usually required three instructions, but now only one instruction is needed to complete. Target workloads that this technology can be applied to include image recognition, image segmentation, speech recognition, language translation, and object detection.