Tech giant Microsoft has pursued dozens of partnerships with health systems over the past few years, with these focusing mainly on fine-tuning AI models to better serve busy clinicians.
Last month at the HLTH conference in Las Vegas, I stopped by Microsoft’s booth to learn a bit more about how the company’s health system partners are deploying AI.
UW Health
UW Health, the University of Wisconsin’s health system, teamed up with Microsoft to advance AI in the imaging space. The partners are creating AI applications that can help radiologists better interpret medical images — as well as have an easier time with disease classification and report generation.
“The very first use case is centered around chest X-rays — can we take an AI model and go directly from imaging to reports? Chest X-rays are only the beginning. Worldwide, chest X-rays are the number one acquired imaging exam — and if you look outside the U.S., the vast majority of those are never actually interpreted by anyone with normal imaging experience. So there’s a massive need out there,” said Richard Bruce, vice chair of informatics at UW Health.
Providence
Providence has collaborated with Microsoft to build a whole-slide foundation model for digital pathology, said Carlo Bifulco, chief medical officer of the health system’s genomics division.
The partners have teamed up to develop a machine learning model that is pretrained on Providence’s real-world pathology data. The model can capture patterns across entire slides, enabling improved mutation predictions and more precise cancer subtyping, Bifulco explained.
“The next step is going to be a conversational step, where [clinicians] can have conversations with this big universe of images. So as a pathologist or as an oncologist taking care of a patient, you will be able to have a chat with the images that have been acquired from the patient — not only with the images, but also with their clinical datasets,” he remarked.
Stanford Health Care
Stanford Health Care is scaling up its use of Microsoft’s Dragon Ambient eXperience (DAX) listening tool. The AI-powered tool creates clinical notes by listening to patient-provider conversations and generates structured documentation in real-time.
Nearly 500 physicians currently use the tool, and Stanford is looking to increase that figure to 1,000 by the end of 2024, according to Christopher Sharp, the health system’s chief medical information officer. Stanford began deploying the tool among its primary care physicians, but it has since expanded its DAX program to include specialties such as gastroenterology, neurology and orthopedics, he noted.
In the future, DAX will help Stanford clinicians with additional workflows beyond just clinical documentation, Sharp pointed out.
“We look at things like the opportunity to tee up the order that’s been spoken and then push it back into EHR. It will say, ‘It sounded like you needed a chest X-ray — here’s your order.’ We also think it will start to influence diagnosis as we go forward,” he said.
Northwestern Medicine
Northwestern Medicine has also deployed Microsoft’s DAX listening tool. The health system deployed the tool mainly among its primary care physicians, along with a “smattering” of specialists, said Kali Arduini Ihde, director of innovation at Northwestern.
Overall, Northwestern physicians using DAX are spending 30% less time on documentation so far, she stated.
“When we gave them the tool, we said, ‘Do what you want with your time.’ If you save time and you want to go home earlier to be with your family, great. If you want to see more patients because now you’re freed up, go for it. And the majority of them gave the time back to their patients,” Arduini Ihde noted.
Photo: Gerasimov174, Getty Images
