S6 Ep7: The Reading Room Podcast: Current Perspectives, Challenges and Possibilities with AI in Breast Imaging, Part 1

What excites you about the potential of AI in breast imaging, and what have you seen thus far in practice and or research?

So I can certainly talk to the research angle of this.

In Europe, it has been very clearly shown to reduce the time spent interpreting and the need for a second reader.

In Europe, every mammogram typically had been read by two radiologists.

And now the AI is a very effective second reader.

And in fact, improves cancer detection more than traditional double reading.

And so that really reduces the workforce needs and also improves the performance of mammographic interpretation.

It's harder to generalize that to the United States since we don't do double reading, but we'll talk more about that, I'm sure.

And then I think the other big issue that we're still tackling is that right now it's very inefficient to try to do a full risk assessment on every patient.

We've had guidelines since 2007 from the American Cancer Society saying that we should be doing screening MRI in women who have a lifetime risk of breast cancer more than 20%.

But to try to calculate that or estimate that from, for example, the Terracusic model, which is the most accurate, is a time-consuming and manpower-consuming process that takes at least sort of 10 minutes, and we don't routinely collect all that information that would be necessary.

And so AI actually, just by looking at the mammogram alone, can estimate risk at least as accurately as these risk models.

And so there would be no manpower really involved in that.

And I think that's a very exciting potential use.

I know Manisha has used that in their practice and written about that, so I'm sure she can speak more to that as well.

What excites me about AI is its potential to improve breast imaging across the entire patient journey, from cancer detection to personalized risk prediction, as Dr. Berg said.

We've moved beyond the retrospective studies and simulation studies to prospective studies and real-world evidence of benefit.

For example, the MASAI trial, a large randomized trial in Sweden, showed that AI-assisted mammography improves cancer detection and sensitivity without compromising specificity.

And I think that cancer detection is just the beginning.

I echo, you know, Dr. Berg and Dr. Ball's sentiments.

And, you know, I look at AI, you know, I'm incredibly excited about its prospect in breast imaging across all modalities, ultimately.

I know today we're focusing more on the mammography portion.

But, you know, I see a world and I know I'm kind of a dreamer in this sense, but I see a world where we move from, you know, just from sole algorithms to an entire workflow.

So let's say, for example, AI, you know, identifies a potentially missed cancer.

Then we have it flagged for review by a radiologist or breast imaging specialist.