NEW YORK – With two US Food and Drug Administration-cleared platforms now on the market, digital pathology is poised to carve out a significant space in patient care.
Some clinicians question, though, whether the near-term benefits of digital pathology justify immediate investment in the technology, or if labs are better off waiting for more advanced platforms and applications that are expected from the technology further down the road.
Digital pathology uses scanners to create digital images of pathology slides of biopsied tissues, which can then be shared and viewed as digital files. In the short term, the technology has the potential to enable easier, faster, and less expensive sharing of pathology slides while also streamlining slide storage and management.
Longer term, digital pathology provides the foundation for the application of artificial intelligence to pathology, allowing software to assist pathologists in their reads, or possibly even to take the lead in certain analyses. A number of companies are currently developing AI-based pathology applications, though none of these applications has yet received FDA clearance.
Philips was the first company to go to market with an FDA-cleared digital pathology system. The firm received clearance for its IntelliSite Pathology Solution in 2017. Recently it announced a pair of major deals involving the technology.
Alverno Clinical Laboratories, which manages 33 hospital labs and provides pathology services to thousands of physician offices throughout the Midwest, said in May that it would be moving fully to digital pathology on the Philips system.
In June, Laboratory Corporation of America and New York's Mount Sinai Hospital said that they were working with Philips to establish the Mount Sinai Digital and Artificial Intelligence Enabled Pathology Center of Excellence using the IntelliSite Pathology platform. LabCorp previously installed the Philips digital pathology system in four of its laboratories in 2018.
Meanwhile, in May, Danaher's Leica Biosystems received FDA clearance for its Aperio AT2 DX digital pathology system, bringing a second cleared platform to market.
While uptake of digital pathology is still in its early stages, "the wheels are turning," said David Rimm, professor of pathology and medicine at Yale School of Medicine. "A lot of places are saying, yeah, we're not digital yet, but we will be in six months, or a year, or a year-and-a-half."
He said that the future potential of digital pathology, more than its current advantages, was driving this shift within the field.
While the improved ease of accessing and sharing slides provided by digital pathology is nice, "it's hard to quantify how big a deal that is and how much money that is saving you," Rimm said.
AI-based applications, on the other hand, have the potential to drive tremendous cost and labor savings, he said. He cited the example of pathology firm Paige.AI, which is developing AI approaches for evaluating digital pathology slides.
In March, the company received FDA Breakthrough Device Designation for a prostate cancer pathology tool that Rimm said could substantially increase the efficiency of pathologists working in this area. Rimm is a member Paige's advisory board.
The company's software uses AI to detect cancer cells in prostate biopsy samples, which Rimm said could be useful as part of quality control processes checking pathologists' work. More exciting, he added, is the potential for the software to identify slides with no sign of cancer that don't need to be looked at by a pathologist.
"That saves [labor], that saves workloads, and that is something we have been waiting for, to have [digital pathology] save us money," he said. "If that gets FDA-approved and we bring it on, that is going to be a large workload [pathologists] won't have to do anymore. Sixty percent or 80 percent of negative core [prostate] biopsies won't have to be looked at by a human. They'll just be processed by the machine, they will be marked as negative, and off we will go."
Rimm noted that this practice has been common in cytology for years now.
"Around 20 percent of our cytology cases are never looked at by a human, and that saves us money" he said. "It's just now coming to histology. Prostate will probably be first, and maybe breast and colon will be soon to follow."
"It's the AI companies that are driving this, but for them to be successful, there has to be digital pathology," he added.
A number of firms beyond Paige.AI are developing AI-based tools for digital pathology, including Boston-based PathAI, which closed a $60 million Series B round in April; Columbus, Ohio-based Deep Lens, which closed a $14 million Series A round, also in April; and Israel-based Ibex Medical, which is developing AI-based software for aiding analysis of prostate core needle biopsies.
Tech industry heavyweights like Google are also exploring AI-based applications for digital pathology, as are a variety of academic groups.
Like Rimm, Nicolas Cacciabeve, a pathologist with Rockville, Maryland-based Advanced Pathology Associates, said that while digital pathology itself offers an advantage in terms of convenience and efficiency, this is "just a drop in the bucket" compared to AI applications, which he described as "the real disruptor."
Even so, Cacciabeve said the advantages of digital pathology as it currently exists are potentially substantial.
"For producing an anatomical pathology report, the single largest cost is the labor component of the pathologist," he said. "If you can make the pathologist 15 percent to 20 percent more cost efficient – and by that I mean that in the same number of working house they can sign out an additional 15 percent to 20 percent of cases because they are presented to them more efficiently – then you have saved 15 percent to 20 percent on the highest-cost portion of the process."
Additionally, he said he believed digital pathology would improve accuracy by allowing clinicians to refer slides more quickly to experts and would improve precision by allowing multiple pathologists to more easily look at a slide at the same time.
Advanced Pathology Associates currently uses a Roche Ventana system to digitize control slides to allow for easier sharing among pathologists at different hospitals. The practice also participated in studies that went into Philips' FDA submission for the IntelliSite Pathology system, Cacciabeve said.
He said the practice has discussed moving over to digital pathology but has held off thus far due to the upfront investment that would be required.
"We're a relatively small practice with 15 pathologists and seven hospitals," he said. "There are cost savings to using digital imaging, but the current cost of buying the imagers is still higher than what we as a medium-sized practice can afford."
Cacciabeve said, however, that he expected equipment costs would come down as more firms brought FDA-cleared systems to market.
He added that he anticipated that imaging firms would structure pricing to reduce upfront capital costs as they move to market the technology to smaller lab groups.
"I think the current marketing has focused on larger institutions like universities and larger practices, and for them it is not a problem to put down the capital," he said. "But there is a whole other market out there that is probably a little more capital averse. And so just as it happened in the clinical lab, there will be smart marketing people who will come in and say, 'We're not going to sell you the instrument anymore. We're going to lease it to you and charge you a per-click cost, similar to a reagent rental [model].'"
Rimm noted that in addition to upfront equipment costs, labs will need to consider the costs of storing the digital files.
"Do we store everything?" he said. "That's going to be expensive. But does the fact that we have [slides] stored digitally mean we can throw out our glass and save money on archiving? There are a lot of variables that are yet to be determined that will feed into the total cost."
According to a 2019 survey of pathologists by consulting firm Diaceutics, capital and ongoing costs are moderate barriers to entry for labs considering adopting digital pathology.
More pressing, though, according to the 35 pathologists the company surveyed, are questions around how to validate digital pathology systems and how to implement the new workflows as well as concerns about regulatory compliance.
Frank Policht, a senior director at Diaceutics, noted that while the College of American Pathologists this year released guidelines on the use of digital pathology for quantitative image analysis (QIA) of HER2, institutions have largely been on their own in terms of validating assays using the technology.
"Sometimes you need a couple of institutions to really break the ice, so to speak, and establish a model [for validation and implementation], because without [that] kind of a framework people can feel lost," he said.
Policht highlighted as another key issue the survey's finding that 38 percent of samples evaluated using QIA needed to be prepared again for manual evaluation by a pathologist because of problems with specimen quality or slide preparation. This is more than twice what is typically considered an acceptable rate, he noted, suggesting that this might also be a barrier to adoption.
Robert Boorstein, a pathologist and medical director at Lenco Diagnostic Laboratory, headquartered in Brooklyn, was more skeptical than Rimm and Cacciabeve that digital pathology would see significantly accelerated uptake — at least until the emergence of AI-enabled applications.
As it currently exists, digital pathology "doesn't help the pathologist do their job that much better," he said. "And there is no inherent, obvious savings from simply using the technology."
"It may be that ultimately the AI solutions drive this," Boorstein said, but he added that he believed it was probably "not a great investment to invest in [digital pathology] hardware now" in anticipation of future applications.
"As a general rule of thumb, I don't think it makes sense in the clinical setting to buy hardware where people say, 'You could have this great application in two or three years when it gets FDA approval,'" he said. "The cost of hardware is always going to go down."
Rimm suggested, though, that labs could struggle to catch up if they haven't laid the groundwork for adoption of digital pathology before the first AI applications come to market.
"If you want to be a competitive site, then you need to be ready when the AI comes out," he said. "You can't say, 'Oh, I'm going to wait for AI,' and then when it comes out say, 'I think I'll go digital now,' because then you're two years behind the curve, because that is how long it takes to go digital."
"So I think many of us are thinking, OK, AI is pretty close, let's get ourselves digital so we can take advantage of it when it hits," he said.
Rimm added that he believes improvements in digital pathology platforms will also drive uptake.
Specifically, he said that the Leica Aperio AT2 DX system uses a more open data format than the Philips platform, which he said could make it more attractive to customers, particularly hospital systems.
"The Philips scanner [uses] an image format that we can't read [without Philips software]," Rimm said, adding that the file format used by the Leica system is compatible with a wide range of software tools.
"In my opinion, that has scared people away [from the Philips system]," he said. "I know a couple of places that have tried it and then said, 'No thanks,' and a lot of the reason for that was the closed system."
Margaret Horton, head of marketing, digital and computational pathology at Philips, said that the choice of file format was guided by the requirements of the FDA regulatory process.
Rimm said that Yale is beginning to ramp up its investments in digital pathology technology with the goal of being fully digital in three years.
A number of larger labs, he noted, have already invested significantly in the technology.
"They are scanning a certain percentage of their slides because they see it's going to be a competitive advantage to be digital," he said. "And I agree with that. It may become something that you advertise — that you are digital and you send [customers] digital images of every slide. And that may make people send their specimens to you as opposed to some other competitor."
Alverno plans to be fully digital by early 2020, said President and CEO Sam Terese, noting that the company currently has eight Philips scanners in place with plans to add another four toward the end of the year.
As the Diaceutics survey suggested, validation has proven a somewhat challenging part of the process.
"It's a little harder to validate because there is no one to go talk to about how to do it," Terese said. "We've had to create [a validation process], and I hope that makes it easier for others."
Like Rimm and Cacciabeve, Terese said digital pathology will become "really interesting" several years in the future when AI and computer-assisted diagnosis applications become available.
However, he added, the immediate advantages are also substantial given the many different sites Alverno services.
"We generate around 1.1 million to 1.2 million slides each year, so that translates to about 5,000 to 6,000 slides a night," he said. "Each and every day we have to send those slides to numerous locations throughout our network. If you think about what digital pathology is going to do for us, basically all that work will go away. No longer will we be putting thousands of slides in cars and driving them through all sorts of weather and whatever else."
That will allow the company to cut its courier costs dramatically while also speeding up turnaround time by what Terese estimated would be an average of around 24 hours.
He said the digital system will also enable easier sharing of slides, allowing, for instance, "a pathologist who doesn't have a certain expertise to rely on a partner or someone somewhere else to kind of fill in the blanks for them. So, I think we will improve the health for the overall community that we serve."
LabCorp announced last year that it would install the Philips IntelliSite Pathology system in four of its labs. Under its collaboration with Philips and Mount Sinai to establish the Mount Sinai Digital and Artificial Intelligence Enabled Pathology Center of Excellence, LabCorp will help integrate digital pathology throughout the Mount Sinai system. Additionally, Mount Sinai pathologists will provide expert consultations when needed on samples from LabCorp customers.
"We have … key opinion leaders of national and international reputations, and we already receive on a routine basis [specimens] from patients both inside and outside New York, and digital pathology is a tool that will ease that extramural clinical services that we offer the community," said Carlos Cordon-Cardo, chairman of Mount Sinai's department of pathology, molecular and cell-based medicine.
Cordon-Cardo and his colleagues are also developing AI-based tools for applications including making mitotic counts in cancer specimens and grading prostate and breast cancer specimens.