Envisioning a Future Shaped by Radiologists and AI

“Our current processes and technologies just aren’t serving us, there’s a massive amount of volume of information coming into the system and our turnaround times are increasing,” said Dr. Kottler, associate chief medical officer, clinical AI at Radiology Partners, headquartered in El Segundo, CA. 

As a result, she said “Other people are trying to solve our problems. But we need to be the ones creating our own vision.”

The massive influx of information coming into radiology includes not only AI, but also data from molecular imaging, radiomics, proteomics, genomics, and wearable devices. “I suggest the right person to manage this onslaught of information is the radiologist. An integrated information and integrated communications platform and intelligent connections between humans and technology will help us get there," Dr. Kottler said.  

“The goal is an expert radiologist who is informed on how the AI works partnering with a transparent and explainable AI system,” she said. “Together, they’re better than either alone. That’s augmenting human intellect.”

She shared several examples from the more than 200 million AI-interpreted reports, demonstrating how AI improves radiologist accuracy and vice versa. 

“Radiologists and AI have different biases and interpret images differently,” she said. “When they don’t overlap is where you get the benefit.” 

Just as a radiologist and resident combined provide an ideal, but un-scaleable, exam interpretation, radiologists and AI together without bias can provide an outsized impact, she noted. 

“Bias at the point of care can absolutely be mitigated,” she said. “The radiologist should be informed as to where AI tends to make mistakes and where it tends to be right. You also need AI systems that are transparent and explainable.”

When it comes to human-to-human connections, Dr. Kottler stressed the importance of her operations committee, a diverse team who take ownership and a bias toward action. Clinicians are important components of that team, along with “cat herders,” able bring together disparate functions of an organization.

Dr. Kottler said that unlike one-off AI tools currently in use, new foundational models, or generative AI, will fundamentally change how radiologists work.

“Generative AI will be able to look at the entire chest X-ray and draft a report for you,” Dr. Kottler said. “Instead of dictating and reviewing, AI will tell us what it sees and provide an interactive report while automatically performing administrative tasks.”

She challenged her colleagues to imagine the tools that will help them and the workflows they would automate.

“Embrace this change and be a leader of change within your organization, because only by accepting it do we get to create our vision of the future,” she said. “We should be the ones defining our own future. We know the workflows. We need to create the tools that will change the practice of radiology.” 

Access the plenary session, “The Only Way to Predict the Future Is to Create It,” (S6-PL01) on demand RSNA.org/MeetingCentral