In Conversation with Matt Lungren

In this issue of Further's Own the Unknown? LinkedIn newsletter, we will highlight our insightful conversation between Dr. Matthew Lungren and Further’s Keith McCormick , recorded on January 8th. Our next conversation will be very soon. We will be live with author and AI ethicist Olivia Gambelin on February 12th.?

Matt has multiple roles. At the top of the interview, he described his Microsoft role as “bidirectional.” That includes working with Microsoft’s technology partners to identify how solutions can be purpose-fit for healthcare, and presenting healthcare challenges to the research and technology teams to see if that can inspire the development of innovative solutions. “It’s pretty each to lose sight of the reality of healthcare as it is currently practiced”. For instance, he reminded the audience that modern healthcare systems still use fax machines. To stay connected, Matt still practices medicine, and he still works with students at Stanford. Matt was a co-founder of the Stanford Center for Artificial Intelligence in Medicine and Imaging (AIMI Center).

The Junior High School Dance and Expert Collaboration in Healthcare

When asked about the AIMI Center, Matt shared a wonderful metaphor.

There’s this…junior high school dance where the computer scientists are on one side and the clinicians are on the other. They want to dance, but they don’t know how to talk to each other. The AIMI Center was set up to bridge that gap.

We also shared a bit of that part of the conversation in this video clip on LinkedIn. In a second clip, he shared how data scientists and subject matter experts, like doctors, should spend their time together. This resonated with Keith as it mirrors the importance of client expertise in our consulting engagements, which are highly collaborative.

Computer vision, contrast agents, and medical imaging

Keith asked about technology that can avoid the use (or reduce the amount) of contrast agent for medical imaging, which has to be ingested by the patient. This use of this technology is already several years old, and it continues to evolve. But Matt noted that it is a striking example in that both hospitals and patients can immediately see the benefit.

It’s about shortening the time a patient spends in an MRI machine while maintaining diagnostic accuracy. This not only improves patient comfort but also aligns incentives for hospitals by increasing throughput. The really beautiful thing about some of that work is that you’re helping the hospital and the patient at the same time, and often times that doesn’t always line up like that. The patient’s like, ‘Cool. I only have to be in this magnet for half the time,’ and the hospital’s like, ‘Cool. I can actually get through my backlog of patients who need imaging.’

Systemic inefficiencies and addressing physician burnout with AI transcription

One of the highlights of the conversation was discussing medical transcription. As Matt points out in this video clip, doctors have actually postponed their retirement because technology has reduced the arduous and exhausting requirement of constant paperwork. Many have experienced this as patients, as he describes.

You walk into a doctor’s office, and instead of them typing notes while you’re talking, the conversation is transcribed. At the end of the visit, the note is generated, excluding irrelevant details, giving the clinician a chance to edit and sign off. Over 500 health systems have adopted this technology, saving doctors up to 50% of their time. And that’s just the start, right? If we can have physicians that end their workday on time, are delighted, and their patients are delighted, we’re seeing this as a massive hit.

The theme of inefficiencies, and the frustrations they can cause, came up again later in the interview.

The U.S. healthcare system is like having Ferraris stuck in LA traffic. You have these amazing tools, but the system isn’t designed to fully leverage them. You’re still not getting through to your destination any more quickly than with a Pinto. We can’t just throw concrete onto the same processes forever and expect everything to work out. There’s enough incentive now, probably more than ever, to start moving the needle.

There was another aspect of medical transcription discussed. Since the primary motivator for transcription was originally billing-related paperwork, Keith asked how well-suited the transcription data was for other use cases, like predictive medicine.

We’re moving towards a space where data becomes dynamic. For example, a note generated for billing can also be adapted to a seventh-grade reading level for patients or refined for referring clinicians. Formatting things starts to become less of a challenge because we can adapt the same content to serve multiple use cases. This kind of flexibility is what’s exciting about the infrastructure shift we’re seeing in healthcare today.

AI Ethics and Medical Education

AI Ethics is a topic that we will be discussing in depth with Olvia Gambelin in our next thorough leadership interview, but Keith also asked Matt about it. His answer drew upon his experience working with students at Stanford.

You can’t overemphasize the importance of addressing bias. This includes biases in the technology, automation bias, and confirmation bias. These challenges must be front and center in any education effort introducing AI to healthcare. It’s critical. And I think any education effort that is going to bring AI to a new domain, particularly in healthcare, this has to be at the top of the list of things that get covered.

His answer was also featured in a recent video clip.

Coming Soon

Don’t miss out. Our next interview is in just a few days on February 12th. Also, Nicolas Decavel-Bueff will be presenting his masterclass Hands-On Introduction to Customizing Large Language Models in Las Vegas the end of this month. And the leader of our AI and Data Science team, Cal Al-Dhubaib , is a confirmed speaker at ODSC East.