Eric Topol

So we are not using AI as we should.

We shouldn't be charging patients.

And we're only using the first of the three different types of AI to get the superhuman interpretation of the mammogram.

Right.

Well, the medical imaging is the most advanced part of medical AI because it's been incubating and progressing for several years now.

And so now for most medical images, whether it's an x-ray or a mammogram, CAT scan, PET scans, whatever type of scan, and also things like the retina, the cardiogram, pathology slides,

We now have very strong evidence that AI can interpret the image better than expert physicians, which is a big step forward for AI.

So it enhances the accuracy.

It doesn't mean that it shouldn't be reviewed by physicians, but it sees things that we can't see, superhuman eyes, if you will.

And so the problem is, Manoush, we haven't leveraged that at all.

And the only thing that barely we're getting at are things like just a piece of the mammogram and not having that appropriately embedded as part of the procedure rather than putting a burden on patients for cost.

And we're not doing it for most of the things that we know it has remarkable capabilities for accuracy.

So this is the area that has come along the most.

But of course, it's been worked on for several years.

The thing that we don't talk much about are diagnostic medical errors.

And according to the National Academy of Medicine, almost all of us will experience at least one in our lifetime.

We know from a recent Johns Hopkins study that these errors have led to

800,000 Americans dead or seriously disabled each year.

So this is a big problem.

The question is, can AI help us?