Paige.AI nabs $25M, inks IP deal with Sloan Kettering to bring machine learning to cancer pathology

Next Story

Lightmatter aims to reinvent AI-specific chips with photonic computing and $11M in funding

Artificial intelligence has become one of the key weapons in the fight against cancer and the many forms and mutations that it takes, and today a startup is coming out of stealth and announcing funding and a significant data deal as it seeks to build an AI system specifically to help understand one aspect of the treatment cycle: cancer pathology.

New York-based Paige.AI — an acronym for Pathology AI Guidance Engine — has closed $25 million in Series A funding and has signed a deal with the Memorial Sloan Kettering Cancer to have exclusive access to its 25 million pathology slides (one of the biggest repositories in the world) as well as its intellectual property related to computational pathology.

The company said that it plans to focus first on breast, prostrate and other major cancers, and expects to partner with other medical centers beyond MSK, as well as commercial labs and pharmaceutical companies as it grows and develops its applications.

The funding, meanwhile, is led by Jim Breyer of Breyer Capital, along with other, unnamed investors that “wish to remain anonymous”, the company tells me, and appears to be the first funding ever announced by Paige. MSK received equity as part of the license, but MSK isn’t a cash investor.

“Paige.AI is poised to become a powerhouse in computational pathology and an undisputed leader among thousands of healthcare AI competitors,” said Breyer in a statement. “Today, we take a major step forward in harnessing machine learning and more fully realizing its promise for cancer diagnosis and treatment.”

Notably, the startup was essentially hatched inside of MSK itself.

One founder who is now the CEO, Dr Thomas Fuchs, is known as the “father of computational pathology” and is the director of Computational Pathology in The Warren Alpert Center for Digital and Computational Pathology at Memorial Sloan Kettering, as well as a professor of machine learning at the Weill Cornell Graduate School of Medical Sciences.

The other co-founder is Dr David Klimstra, who is chairman of the department of pathology at MSK. The company’s chairman, Norman Selby, has been the chairman of two other medical information businesses, Real Endpoints and Physicians Interactive, in a long list of other roles.

Both are keeping their existing roles as they also build out Paige.AI, the company confirmed to me.

Using AI to tackle the challenge of cancer — with all its millions of permutations and variations — is not new. As some examples, researchers have done work into using AI to detect colorectal cancerbreast cancer and lung cancer earlier. And tech companies like Microsoft are committing resources as well to the prospect of using AI to find cures.

Paige focuses specifically on cancer pathology — that is, reading tissue samples of people who already have the disease, and have had biopsies and other tests in order to determine what kind of cancer is being dealt with, how advanced it is, whether it’s a primary or secondary, and other details: these details become essential to get right so that a patient can be suggested more targeted treatments (another area where AI and other advanced tech is being used to craft more nuanced approaches specific to the person being treated).

AI is uniquely positioned to help with how pathology is approached: many of the techniques that are used today are manual and developed over 100 years ago, Paige points out, and have not kept pace with being able to process and “read” the amount of data that we are now able to obtain from a patient. To find a team of people who could do the work of an AI system in reading and processing all this would be more costly and might not even be possible.

“Patients deserve and need an accurate diagnosis as quickly as possible, yet our current methods are time-consuming, expensive and subjective,” said Klimstra in a statement. “The field is ripe for innovation and we are confident that Paige.AI will aid pathologists in detecting disease better and faster. With computational pathology, pathologists can redirect their efforts toward more sophisticated tasks, such as integrating histologic findings with other diagnostic analyses.”

There are two sides to what Paige.AI is attempting to do. One is to ingest a large amount to data to train its AI platform about as many examples of cancers, diagnoses and outcomes as it can. MSK had already digitised 5 million of the 25 million pathology slides that will be donated to Paige for its efforts, and Paige has committed to digitising the remaining 20 million.

The other is to help “teach” that system some of the smarts from existing researchers to put that data to work in the best way possible.

Fuchs compares the task to autonomous car systems, which are only valid if they can “think” like humans to recognise things in complex arrangements and under different conditions. He believes computational pathology is that so-called “missing link” in cancer diagnoses.

And in that vein — similar to a lot of autonomous services, in fact — it’s important to note that Paige.AI does not see itself as a replacement for humans (and in its case human pathologists), but as a tool to help them make better decisions for their patients.

“If you’re teaching a self-driving car on a closed course, anyone can label a tree or a sign so the system can recognize it,” said Dr. Fuchs in a statement. “But imagine the additional guidance that car would require to navigate New York City. The same is true in a specialized medical domain like oncology. You must have both massive data sets and specialists with decades of training to ensure that the computer models are up to difficult tasks. We will enable computational pathology to expand at the scale necessary to achieve intelligent, quantitative clinical models – and facilitate widespread adoption of digital pathology,” Fuchs added.

Fuchs has an interesting background that also points to how solving problems in the world of cancer are not unlike other challenges in detection.  Before joining MSK, he also worked on the Mars Rover project at NASA, where “some of the same algorithms used to identify terrain on Mars are useful in differentiating cancerous from benign tissues on slides.”

This is not the first startup to be spun out of MSK. Drug maker Juno Therapeutics is another; it is now getting acquired for $9 billion by Celgene.

Featured Image: Daniele Carotenuto Photography/Getty Images