Intel and GE Healthcare Partner to Advance AI in Medical Imaging

GE Healthcare used Intel® Vision technology to develop an AI algorithm that can help medical staff triage potentially life-threatening cases faster.

At a glance:

  • GE Healthcare has introduced a first-of-its-kind AI algorithm that is embedded on the X-ray imaging device to help raise user productivity, expedite time to diagnosis, and enhance patient care.

  • The Intel® Distribution of OpenVINO™ toolkit improved algorithm performance, dropping pneumothorax inferencing time and the time to analyze an X-ray from more than three seconds to under one second.1

  • AI-enhanced x-ray devices can flag critical cases on the device and send to radiologists for immediate triage.

  • For AI solutions to be adopted, they should integrate with existing workflows and help users perform their jobs more effectively.

author-image

By

New X-Ray Intelligence Promises Improved Patient Care

How could artificial intelligence (AI) be used to advance medical imaging and improve patient outcomes? That was the question GE Healthcare sought to answer by immersing itself in hospital workflows. They interviewed radiologists and technologists and observed operations across the hospital to determine how staff uses X-ray equipment.

Next, hospital staff were asked to identify which of X-ray’s many uses would benefit most from AI. The answer was clear: they wanted AI to help improve the handling of cases involving critical findings to better treat patients. Specifically, they sought assistance with those conditions that, while not common, could lead to life-threatening consequences if overlooked.

Another pain point was the current first-in-first-out workflow, which commonly resulted in long turn-around times (TATs), even among cases designated as emergency or stat. The practice created a queue of stat cases with little clarity about which were to be addressed first. Users wanted a solution that would help ensure that the right person evaluated the X-ray at the right time, enabling the best possible patient care.

GE Healthcare’s Critical Care Suite Embeds AI on the Device

In response to the needs of its customers, GE Healthcare developed its Critical Care Suite, a set of AI algorithms built to detect critical findings on a chest X-ray. Available on Optima XR240amx systems, the solution represents a significant step forward for the analysis of X-rays, which account for about 50 percent of medical imaging today.2

“What’s different about what GE Healthcare is doing is it’s a first-of-its-kind AI algorithm that’s embedded on the imaging device,” said Todd Minnigh, CMO X-Ray, GE Healthcare. “So, the thing that’s actually capturing the images is also doing the processing. It’s not in the cloud and not on a server downstream somewhere. It can detect and enable providers to prioritize critical conditions, so cases can be more quickly flagged.”

Running the algorithm at the point of care on the same Intel® processor-based systems conducting the X-rays enables the hospital to identify key findings faster. AI findings of potential critical conditions are sent to the radiologist at the same time as the original image is sent to the picture archiving and communication system (PACS). Plus, the ability to identify and flag quality issues in real time allows the technologist to determine if an image should be repeated or reprocessed while the patient is still in the lab.

All of this can mean quicker access to results for anxious patients, easier reprioritization of workloads for busy medical professionals, and potentially improved patient outcomes.

GE Healthcare sought to test the solution by addressing one especially challenging use case that uses X-rays to examine for pneumothorax, a life-threatening, difficult-to-detect condition in which air or gas has entered the cavity between the lungs and the chest wall, causing the lung to collapse.

What’s different about what GE Healthcare is doing is it’s a first-of-its-kind AI algorithm that’s embedded on the imaging device. It’s not in the cloud and not on a server downstream somewhere. It can detect and enable providers to prioritize critical conditions, so cases can be more quickly flagged.

Todd Minnigh, CMO X-Ray, GE Healthcare

Intel Helps GE Healthcare Accelerate Pneumothorax Detection on the Optima XR240amx X-Ray System by More Than 3x

Today, most AI solutions targeting hospital workflows are located in the cloud or on a hospital server. With the Critical Care Suite, GE Healthcare wanted to minimize cost, shorten installation time, and reduce security vulnerabilities.3 Every second counts, so processing and intelligence were located in the imaging device itself, thereby avoiding unnecessary delays. It also negated the need for replacement or supplementary infrastructure.

The goal of the Critical Care Suite is to optimize the frontal chest and lung field position in X-rays while expediting delivery of the results of the pneumothorax inferencing once the image has been captured. Moving the compute intelligence to the machine level would allow for fast workflows and enable radiologists to rapidly process results for anxious patients looking for peace of mind.

For support, GE Healthcare turned to Intel and its computer vision tools. Intel shared the commitment to improve the speed of GE X-ray devices and deliver higher-quality X-ray imaging to enhance patient care and outcomes. Intel helped optimize the Critical Care Suite algorithms using the Intel® Distribution of OpenVINO™ toolkit. The toolkit provided computer vision and deep learning inference tools, including convolutional image-based classification models optimized for the Intel® processors used in GE Healthcare imaging systems.

The move to the Intel® Distribution of OpenVINO™ toolkit improved performance across all models, with the pneumothorax model receiving the most benefit with inferencing time dropping from more than three seconds to under one second.1 Pneumothorax inferencing on the Optima XR240amx X-ray system accelerated by 3.3x compared to inferencing without optimizations.1 Intel also helped train the GE Healthcare team to get the most out of its algorithm going forward.

“Key to anyone’s success is you want to pick partners who are on the journey to help you be successful,” said Katelyn Nye, X-ray Global Product Manager, Artificial Intelligence and Analytics, GE Healthcare. “And Intel definitely did that for us.”

Also important to GE Healthcare was being able to serve its large install base of Intel-run systems. The AI-powered innovation needed to be available to all through software upgrades or via a clear hardware upgrade path for older systems.

“For AI solutions to be adopted, they should integrate with existing workflows,” said Nye. “You don’t want to add any overhead or burden. GE Healthcare’s approach to building intelligent machines is avoiding any additional steps, workflow, or infrastructure if the task can be performed with what the customer already has today.”

Uncovering Small Issues Before They Become Big Ones

In addition to the AI-based pneumothorax solution, GE Healthcare wanted AI applications that helped technologists further contribute to improved patient care. That commitment led to the development of three additional quality-based algorithms designed to help guide or coach the technologist, regardless of their level of experience.

One example is the new intelligent autorotate algorithm. Technologists have typically been required to manually rotate images to achieve the proper orientation. Estimates suggest that automating that task alone will save technologists approximately 70,000 button clicks a year, or up to three full working days.4

GE Healthcare’s mission is to continue to find X-ray-related tasks that can be automated. In this way, they can unlock data insights faster and free the technologist to move more quickly through the 50 petabytes of data that hospitals now produce each year.5

Working together, GE Healthcare and Intel understand that faster, higher-quality X-ray imaging can lead to more productive staff, better care, and a healthier world.

Download the PDF ›