Optimizing Medical Imaging at the Edge

 

More and more of today’s medical imaging devices, such as CT, ultrasound, and MRI scanners, rely on real-time AI inferencing at the edge to make critical medical decisions while patients are being treated. Intel’s Deepthi Karkada, a deep-learning software engineer, and Ryan Loney, Product Manager for OpenVINO spoke to Hilary Kennedy about recent trends in AI-based medical imaging and how Intel and its partners are helping identify and address the rapidly changing needs of this burgeoning industry.

“Real-time medical imaging at the edge is important because it enables healthcare providers to get results from scans, run inferences, and make decisions about medical care at the patient’s bedside,” says Looney. “Often these results need to be obtained and processed in two seconds or less.”

Computing at the edge is not without its issues, however. Three of the major hurdles Intel and its partners routinely face are: limited memory in low-power devices, binary size, and latency. “Every megabyte counts when you’re deploying on low-power medical devices with limited memory,” says Looney. “Analytics need to be run in as close to real-time as possible.”

“We know that AI and similar techniques are being adopted in the fields of medical imaging,” Karkada said. “These techniques include things like object detection and semantics segmentation. These techniques help radiologists quickly identify issues and result in many benefits. Many of our partners have been leveraging these advancements in these technologies.”

“Intel offers a portfolio of hardware solutions targeted for AI inferencing,” Karkada said. “This includes solutions like the Intel Xeon® processors, core processors, and FPGAs, that our partners have been able to leverage. On the software side, our OpenVINO Toolkit provides accelerated inferencing solutions. These also take advantage of the hardware features, so they’re tightly coupled and integrated.”

Learn more about AI and edge solutions for medical imaging, and other health and life sciences, by connecting with Deepthi Karkada and Ryan Loney on LinkedIn, or read more about Intel’s medical imaging solutions online.

Learn how to optimize a CT model using OpenVINO here: https://github.com/openvinotoolkit/openvino_notebooks/tree/main/notebooks/110-ct-segmentation-quantize

Hear some of our customer success stories here.

Subscribe to the “Health and Life Sciences at the Edge” channel on Apple Podcasts, Spotify, Google Podcasts, or Simplecast to hear more from the Intel Internet of Things Group.

Recent Episodes

In the rapidly advancing field of cancer immunotherapy, accurately modeling the tumor microenvironment (TME) has become essential to improving the predictive power of preclinical drug testing. As immune-modulating therapies surge forward, with over 4,000 immune modulators in development globally, scientists are refining assay technologies that maintain the complexity of patient-specific tumor biology. In vitro platforms…

As cancer immunotherapy continues to reshape treatment landscapes, fine-tuning T-cell responses has become a critical frontier. Recent advances in 3D organoid models and high-content imaging are enabling scientists to closely mimic patient-specific tumor environments—unlocking insights into how T cells behave, respond, and falter under immune checkpoint blockade. With over 4,000 immune modulators in clinical…

As immunotherapy revolutionizes cancer treatment, the need for physiologically relevant preclinical models becomes more urgent than ever. Despite the success of immune checkpoint inhibitors, a large majority of patients fail to achieve long-lasting responses, prompting researchers to explore more complex and predictive assays. The cancer immunity cycle, first described in 2013, remains a central framework…