Building The AI and Human Connection

The combination of human and machine is better than either one alone, according to RSNA President Curtis P. Langlotz, MD, PhD.

“Forming intelligent connections from machine to machine, human to machine, and human to human will lead us to an exciting future. Building these intelligent connections will yield amazing technological innovations, with reduced stress, a more balanced life, and ample time to nurture the most intelligent connections of all: the ones we build with each other,” Dr. Langlotz told the audience in the Arie Crown Theater for Sunday's Opening Session.

In his President's Address, Dr. Langlotz reflected on the theme of this year's annual meeting and explored the advancement of AI in medical imaging. He offered a compelling vision of how radiologists can wield AI tools to their advantage, transforming the patient experience and benefiting health care overall.

“The secret to the success of radiology has always been our embrace of new technologies, from the discovery of the X-ray, through ultrasound, CT, MRI, PACS, theranostics and now these new intelligent systems,” Dr. Langlotz said. “Our unique connection to technology places us in a pivotal role as the health care system adopts these new tools.”

A long-recognized expert in medical imaging AI, Dr. Langlotz said that in the 1980’s, it took four years and a PhD to build a system that could analyze just a few images. “Today, with the right training data, we can build a system in days that has better accuracy than anything that we built back then,” he said.

Dr. Langlotz noted that this acceleration of AI capabilities spurred anxiety among some in medical imaging, yet over time—and with more exposure to AI—the fear has eased for many radiologists. “Anyone who works with AI knows that machine intelligence is different, not better than human intelligence,” Dr. Langlotz said.

Dr. Langlotz offered several key considerations to help achieve the vision of a world enhanced by AI, starting with a call for better access to good data, including for patients. “Let’s ditch the disk. We need to urge our image exchange vendors to enable universal electronic image exchange,” Dr. Langlotz said.

He added that after their care is complete, patients should be able to easily opt-in to donate their de-identified data for research and for training AI algorithms.

Dr. Langlotz emphasized the importance of representation in system design, including having a diverse interdisciplinary team in the room at the outset, participating in design decisions and guiding training data collection to assure the production of fair AI models that provide accurate answers for everyone.

Additionally, Dr. Langlotz said that better understanding of human-machine connections will help avoid ways that AI machines can degrade human performance. “AI models that explain their reasoning and express their confidence (or the lack of it), should be the norm,” he said.

According to Dr. Langlotz, necessary updates to the outdated AI regulatory framework would place more emphasis on what happens after models are deployed. “Model developers should publish ‘model cards,’ like a nutrition label for AI models, to help radiologists determine whether a model will work in their practice,” he said.

Addressing concerns that high-tech advancements might compromise human connections, Dr. Langlotz said the advances are more likely to offer opportunities to delegate the least attractive tasks to the AI models.

“These advances can upskill us all, reduce burnout, and bring better care to underserved areas. And it can do so while we develop richer human connections, like the ones we form in the reading room, the exam room, and at meetings like this one,” Dr. Langlotz concluded.

Access the presentation, “Building Intelligent Connections,” (PL01A) on demand at RSNA.org/MeetingCentral.