Study details

For the study, researchers looked at over 80,000 women between April 2021 and July 2022. Half of the women had two radiologists look at their mammograms without use of AI, while the other half had their scans analyzed by AI and a radiographer, except in cases where the AI generated the highest risk score, which prompted two radiologists to assess the scan.

The researchers found that AI-supported analysis was not only just as effective as traditional screening, but it also led to 20% more cancers being detected. Specifically, the AI-supported analysis resulted in a cancer detection rate of 6 per 1,000 screened, compared to 5 per 1,000 with the traditional analysis.

The researchers also found AI-supported analysis helped reduce radiologists' workload, ultimately leading to them spend 44% less time reading mammograms.

While the study didn't measure specifically how much time was saved by the AI, the researchers determined that if radiologists read around 50 mammograms an hour, it would've taken a single radiologist between four and six months less time to read around 40,000 exams with the help of AI than it would have taken two radiologists alone.

Discussion

While the findings were positive, Kristina Lång, lead author on the study from Lund University in Sweden, said the results "are not enough on their own to confirm that AI is ready to be implemented in mammography screening."

"The greatest potential of AI right now is that it could allow radiologists to be less burdened by the excessive amount of reading," Lång added, saying AI could eliminate the need for two radiologists to read a mammogram, which could allow more radiologists to help patients.

Ultimately, Advisory Board's Julia Elder emphasized that this study does not suggest that AI can replace radiologists. Instead, it demonstrates that working with AI can reduce radiologists' workload and allow them to help more patients.

Stephen Duffy, a professor of cancer screening at Queen Mary University of London, said reducing radiologists' time burden is "an issue of considerable importance in many breast screening programs," but added there could be concerns that AI would over-detect harmless legions.

It's possible that AI could one day help with tumor detection, but a radiologist's job is much more than that, said Laura Heacock, a breast radiologist at NYU Langone Perlmutter Cancer Center.

"If you spend a day with a radiologist, you'll see that how an AI looks at screening a mammogram is really just a faction of how radiologists practices medicine, even in breast imaging," she said. "These tools work best when paired with highly trained radiologists who make the final call on your mammogram. Think of it as a tool like a stethoscope for a cardiologist."

However, Heacock added that with more research, she and her colleagues may end up using AI the way it was featured in the study in the future.

While this study has exciting implications for early cancer detection, Advisory Board's Lindsey Paul noted that it raises the same concerns that arise with other AI tools: How is patient data being protected? How representative were the data inputs used to train the AI model? According to Paul, these questions will need to be answered before provider organizations feel comfortable implementing these tools into regular clinical practice.

But as more AI tools prove to be effective and radiologists demonstrate an interest in using them, Paul suggests that we will start to see more provider organizations implement AI for early cancer detection. However, simply demonstrating effectiveness in one randomized control trial isn't enough — these tools require significant investment.

Moving forward, we'll keep our eyes on the reimbursement and funding landscape to gauge how quickly these tools gain widespread adoption.

For more major trends impacting cancer care in 2023 and beyond, check out our ready-to-use 2023 Cancer Market Trends presentation. (Advisory Board interviews and analysis; Furlong, Politico, 8/2; Christensen, CNN, 8/1)

There are a multitude of existing applications for artificial intelligence (AI) in oncology, but we are increasingly seeing its potential to help cancer providers make important decisions about patient care. As oncology stakeholders fight to address a variety of pressing challenges (such as rising costs, rapid clinical innovation, and unnecessary utilization), these AI innovations have the potential to help. Read on to learn three ways that providers may be able to utilize AI to streamline clinical decision-making and push precision medicine in oncology forward.