NEW YORK – Paige is preparing to offer its portfolio of AI-based cancer detection tests on multiple digital pathology platforms as the firm shifts its focus away from the development of its own software ecosystem and toward making standalone cancer detection tests.
As part of that strategy, the firm announced on Tuesday a deal with Indica Labs to incorporate its diagnostic models into Indica's Halo AP anatomic pathology workflow management system.
Paige announced the strategic shift this month along with the appointment of Razik Yousfi as the company's CEO and chief technology officer. Yousfi joined Paige in 2019 as VP of engineering while the firm was working to translate research from Memorial Sloan Kettering into commercial products. He replaced Andy Moye as CEO.
Yousfi said in an interview that the company plans to offer its algorithm-driven tools for the analysis of whole-slide images and detection of cancer through other companies' software systems including some that have been competitors to the Paige Platform for slide image management and viewing. Those tools include the company's US Food and Drug Administration de novo authorized Paige Prostate test to aid the diagnosis of prostate cancer as well as its various research-use-only cancer detection and differentiation tools.
Yousfi said that bringing digital pathology technologies into a laboratory involves an expensive and difficult process of combining hardware and software systems, and many companies in the space have developed competitors to the Paige Platform. Paige intends to expand its reach to the labs that have chosen different platforms by directing its efforts toward the development of standalone AI-based tests.
"If you wanted to use Paige Prostate before, you would have to be on the Paige Platform," he said.
Because digital pathology lacks hardware and software standardization, that work will involve ensuring that Paige's tests are compatible with other companies' technologies. He noted, however, that the company has long supported the use of its software with a wide range of scanners, and Paige announced in February that the firm had added support for the DICOM image format on the Paige Platform.
Yousfi said that Paige has started talks with other digital pathology platform developers about the technical work that will be needed to integrate Paige's tests with their software and expects to announce some deals in the coming weeks. He declined to name those potential partners but said that Paige aims to work with the top software platforms that are competing in the digital pathology space.
"It is work that we are going to have to undertake, and it is work that our platform partners are going to have to support," he said.
The company also will work with the FDA to update marketing authorizations, he said. The de novo authorization issued in 2021 for Paige Prostate, for example, indicates that it is designed for use with the Paige FullFocus whole-slide image viewing software, which is part of the Paige Platform.
Yousfi said Paige had built the Paige Platform as an ecosystem to run a wide variety of AI-based applications for slide sample analysis, and the company is now considering its options for the system's future. He noted that the company intends to ensure the platform exists in the long term.
"Our customers love it, but we just don't have the necessary bandwidth to continue to build upon it and expand it over time," he said.
As for the company's tests, Yousfi said that Paige has historically developed its clinical applications organ by organ beginning with prostate cancer detection and followed by applications for breast and colon cancers, among others. Last year, the firm launched its Paige Breast Suite to aid the diagnosis of breast cancers and said last fall that it had received FDA breakthrough device designation for the Paige Lymph Node software for the detection of breast cancer metastases in lymph node tissues.
He said, though, that the organ-by-organ approach to test development has involved repeating for each application similar processes of collecting and curating data, creating algorithms to leverage the data, and launching research to collect evidence on the performance of the algorithms for each use case. In recent years, the company began working on its foundational models that can be used to create multi-tissue applications.
"Cancer detection is a task that's very similar across multiple organs, and it is more interesting to build a system that can identify cancer on a wide set of organs before going deeper into each of those different organs," he said.
Paige announced early this year that it had developed its Paige Pan Cancer Suite for the detection of various cancers including rare tumor types and metastatic deposits across at least 17 tissue types. The company said at the time that it plans to seek FDA regulatory review of the multi-cancer test and other tests that are developed using the firm's foundational model.
Yousfi said that Paige is now determining how it can expand its cancer detection capabilities to cover more tissue types, validate the model, and build a single clinical-grade application that could be used by pathologists.
Meanwhile, the company recently launched its second-generation Virchow AI models that are used for the development of clinical pathology applications. Yousfi described those models as AI-based building blocks that customers can use to create customized applications that would be used by pathologists.
The firm developed the Virchow2 and Virchow2G models in a collaboration with Microsoft and trained them using de-identified data from more than 225,000 patients and more than 40 tissue types.
The firm also said last week that it had launched a research-use AI-based tool for the detection and differentiation of cancers. The firm's OmniScreen module was built on the Virchow2 model and trained on 3 million slide images, and it is used for the evaluation of more than 505 genes and the detection of 1,228 molecular biomarkers from images of H&E-stained pathology slides.
The company said that OmniScreen is used in drug development to aid the screening of samples for target genetic mutations and improve patient selection for clinical trials.
Yousfi said that the adoption of digital pathology continues to grow in the US and Europe, although he wishes that the transition was happening more quickly. Healthcare providers and researchers that make the switch are seeing productivity and patient care benefits as they incorporate AI-based tools into their workflow, he said.
By making Paige's tests available on more platforms, the company hopes to make its tests available to more labs and enable customers to develop their own AI-based tools and create a richer ecosystem of AI-based pathology applications.