Artificial Intelligence in Pathology – Has It’s Time Come?
I would argue we are not even close to that point. We are not on a trough but still ascending the peak. We are not at the summit yet. We are all just barely scratching the surface.
We are just 8 years removed from the first devices for whole slide imaging obtaining FDA clearance. We are just now seeing more than 25% of clinical laboratories deploying, adopting, using and not being able to live without whole slide imaging. And this rapid adoption has occurred in the past 5 years with less than 10% of clinical laboratories adopting whole slide imaging prior to this, due to a public health emergency, less enforcement of some regulations and reimbursement streams have helped the matter and helped and sped adoption.
But it is still early days in the grand scheme of things. It reminds me of some early exits in the early 2010s in digital pathology itself among hardware and software vendors. Those products and technologies remain today as part of larger offerings from larger companies.
What will AI do for us? Better, faster, cheaper, more accurate? That’s a tall order? Will the greatest early wins be drug development or cancer prognostics or day to day diagnostics? Likely the former ones rather than the latter ones. And I think that is what we are seeing now. The interest is still pre-clinical, as it was with whole slide imaging in early days. Now becoming mainstream in anatomic pathology laboratories.
But the S-curve of adoption, as it has before in the space will lag behind.
Where are we going? Where have we been? What is the future?
Last month, Tempus purchased Paige AI for $81M, largely in the form of Tempus stock to, as reports mention, for PaigeAI’s datasets from 7 million slides, perhaps most of which were from Memorial Sloan Kettering Cancer Center.
Critics were quick to point out that PaigeAI had raised $241 million and sold for $81 million for these datasets, largely, it appears, although, perhaps there was more to the deal.
Regardless, I began to wonder where on the hype cycle we are. Are we sliding down the trough of disillusionment? Is this as good as it gets? Is this the best pathology artificial intelligence can do? Is the opportunity for a multi-billion dollar market for AI in pathology behind us and never to be reached?
Or, are we still rising towards the Peak of Inflated Expectations? Are we still on an exponential upward slope instead of coming down one? Is this the Peak of Inflated Expectations and when the dust settles we will come up a bit to the Plateau of Productivity some % above where we started but nowhere near where the Peak was.
Critics would argue this does not bode well for pathology AI. From what I saw it seems the hype and excitement and investment and value and returns and so forth are not what we thought they would be.