NEW YORK ─ A recent US Food and Drug Administration authorization that enables Paige to market software that helps pathologists diagnose prostate cancer could lay the groundwork for accelerated adoption of artificial intelligence technology in this discipline, according to digital pathology industry executives.
The de novo authorization of Paige Prostate — the first such green light for artificial intelligence as an aid to pathologists for primary diagnosis — could provide a lift that digital pathology companies generally need to see greater traction in a field that has been traditionally slow to adopt new technology, Paige CEO Leo Grady and others said.
Paige, a New York-based startup, uses AI and deep learning to analyze and learn from the de-identified images of digitized pathology slides, by comparing the data it gathers with health outcomes to create a large database of prostate cancer images. It looks for patterns in biopsied prostate tissue that match those in the database of prostate cancer images.
"When matches are found, the software provides that information to pathologists," Grady said. "It helps them ensure that they don't miss anything, and they have the confidence that they're getting to the right answer for the patient."
Paige Prostate, however, should be used to provided additional, actionable information while pathologists remain responsible for the primary diagnoses, Grady and other diagnostic industry executives noted.
A prostate biopsy is often initiated following an elevated prostate-specific-antigen test result. Most pathologists evaluate a tissue biopsy by viewing slides through a microscope and using their experience, training, and judgement to make a diagnosis. Those who have access to digital pathology technologies can opt instead to analyze digitized slides using software and a viewer, which requires specific equipment, software, and computer storage.
Currently available digital pathology equipment leverages scanners to create the digital images of pathology slides with biopsied tissues and enables electronic sharing of slides through readers. That provides a basis for technological enhancements, such as the application of artificial intelligence to aid in the analysis of slides and subsequent diagnoses.
Nonetheless, the adoption of such systems among pathologists is still in its infancy, and the pathologists need to see clear clinical benefits, which artificial intelligence can bring with "the potential for improvements to the accuracy of a diagnosis," Grady said, adding, "What we are doing with Paige Prostate goes well beyond what can be done with a microscope."
To grant authorization, the FDA evaluated Paige data from a clinical study in which 16 pathologists examined 527 slide images of prostate biopsies that were digitized using a scanner. Of the total, 171 biopsies were cancerous and 356 were benign. The pathologists assessed each slide image with and without assistance from Paige Prostate and saw improvements in the detection of prostate cancer when pathologists used Paige Prostate versus when they did not, the FDA said.
When they used Paige Prostate, pathologists saw a 60 to 70 percent reduction in diagnostic errors, Grady said, adding that the software also enabled nonspecialists "to get to an equivalent level" of diagnostic expertise to that of a subspecialist, and "provided them with additional information that they can choose to incorporate into their diagnostic processes."
Paige said it plans to submit the full data from the FDA study for publication in a peer-reviewed journal.
Quest Diagnostics also sees the potential for clinical advances through the use of AI by its pathologists, and announced a collaboration in May to leverage Paige's machine learning expertise for the discovery of biomarkers that pathologists could use to detect cancers and other diseases.
By applying artificial intelligence to a digital image, Quest is "now able to explore new opportunities" for its pathologists, said Kristie Dolan, general manager of oncology at the firm. Because of the de novo authorization, Quest will begin using Paige Prostate next month to help its pathologists diagnose prostate cancers, and the companies are developing software for pathologists, oncologists, and other healthcare providers to aid in disease diagnosis with solid tumor cancers, such as prostate, breast, colorectal, and lung, the initial focus.
"There have been opportunities to apply digital pathology for some time for secondary consultations and to review cases involving specialty stains, but the ability to use digital pathology for a primary read is fairly new," Dolan said.
In 2017, the FDA cleared the Philips IntelliSite Pathology Solution, a whole-slide imaging system that enables the review and interpretation of digital surgical pathology slides prepared from biopsied tissue. That represented the first time the agency permitted the marketing of a whole-slide imaging system for such purposes.
Since then, numerous companies have raced to develop digital pathology products, including those that incorporate artificial intelligence.
Like many digital pathology products, Paige Prostate should be viewed as a tool that helps pathologists with their diagnoses, Dolan said, adding, "The value of digital pathology lies not in the digitization and sharing of slides but in the use of artificial intelligence and machine learning to aid the pathologist, so we see this as a tool to support high quality, accurate diagnosis."
Other firms seeing the possibility for the greater adoption of digital pathology systems include Israel-based Ibex Medical Analytics and Geneva-based Unilabs, which in July announced it had inked an agreement to implement Ibex’s artificial intelligence-based Galen platform in European pathology labs, starting in Sweden. The Galen platform is CE marked for breast and prostate cancer detection in multiple workflows and was recently granted FDA breakthrough device designation.
Joseph Mossel, the CEO and cofounder of Ibex, said the FDA authorization of Paige Prostate is welcome news for digital pathology firms. "We believe that a robust regulatory framework is essential for healthcare innovation and will help decision-support tools based on AI-algorithms become an indispensable part of the pathologist’s routine," he said.
In March, Philadelphia-based digital pathology firm Proscia announced that it, too, has a strategic partnership with Unilabs, with an aim to accelerate the adoption of artificial intelligence in routine pathology practice. Through the collaboration, Unilabs is validating the efficacy of Proscia’s computational pathology applications and aims to use the technology in its high-throughput pathology workflows.
"The FDA authorization of Paige Prostate represents a significant milestone for the entire field of pathology," said David West, cofounder and CEO of Proscia. "This news, along with the FDA's creation of a new product code for software algorithms to assist users in digital pathology, will serve as a catalyst to rapidly accelerate the already-growing adoption of AI.
"As AI applications in pathology tend to address narrow use cases, organizations must leverage a broad range of solutions to address their full set of needs."
Meanwhile, Abhi Gholap, founder and CEO of digital pathology firm OptraScan, said he believes the FDA's authorization of Paige Prostate will "empower pathologists to bring objectivity in their case interpretations." Further, the authorization plays a role in facilitating future regulatory approvals for AI-based digital pathology products by providing a platform that developers can use to demonstrate "substantial equivalence to this predicate device," he said.
In the future, it will be important to make such tools available for breast, lung, and other cancers, as well, Gholap added.
Indeed, Paige, spun off from Memorial Sloan Kettering Cancer Center in 2017, is at an advanced stage in the development of AI-based digital pathology software to help pathologists detect other cancers, and it plans to soon seek additional US regulatory approvals, Grady said. Last December, the firm obtained CE marking for its Paige Breast cancer detection software, which directs pathologists to suspicious areas for further evaluation of breast biopsies.
Meanwhile, Dolan said Quest and Paige are developing new applications for artificial intelligence across different tissue types to address a variety of clinical needs. Given its scale and breadth, Quest has access to a large quantity of slides associated with early cancers, she said. The firm is sending de-identified slides to Paige so the computation pathology firm can use them to train its algorithms, and expand on the range of cancers for which its software can aid diagnoses.