NEW YORK (360Dx) – The UK this month made a sizeable investment to foster the adoption of artificial intelligence in digital pathology and imaging, pledging £50 million ($64 million) to support five new centers across the country.
While the first task of the centers will be to digitize their workflows to improve efficiency, they also aim to eventually develop new AI-based analytical tools for diagnosing patients. The investment was made through Innovate UK, the country's innovation agency, and will support the centers for the next three years.
Of the five centers, three are focused on the uptake of AI in digital pathology. These include the Industrial Center for AI Research in Digital Diagnostics (iCAIRD) in Glasgow, Scotland; The Pathology Image Data Lake for Analytics, Knowledge, and Education (PathLAKE) in Coventry; and the Northern Pathology Imaging Collaborative (NPIC) based in Leeds.
Two of the centers, the London Medical Imaging and Artificial Intelligence Center for Value-Based Healthcare and the National Consortium of Intelligent Medical Imaging in Oxford are devoted to imaging. A number of companies are taking part in the establishment of the centers, including Philips, Leica Biosystems, Roche Diagnostics, Siemens Healthineers, Canon, and GE Healthcare. All five centers are expected to network with each other as they get underway. The centers are expected to become operational next year.
According to investigators involved in the new centers, the investment reflects not only an interest to improve UK healthcare, as there is general agreement that there is a shortage of pathologists to handle current workloads, but also to secure the British economy, especially given the country's pending exit from the EU, which is expected to happen next March.
"We are in a process of considerable political change," noted David Snead, director of the University Hospitals Coventry and Warwickshire NHS Trust's Digital Pathology Center of Excellence. "There is concern that the economy will take a nosedive because of [Brexit] and the [government] wants to do what it can do to stimulate the economy," he said.
According to Snead, the UK also "sees itself as being in a powerful position" for applying AI in digital pathology, given its background in research and computer science in particular.
Snead is one of the investigators involved in PathLAKE, which received £10 million to apply AI in digital pathology. In addition to UHCW NHS Trust, the effort involves multiple teams across the UK, as well as Philips as an industry partner. Among its objectives are to cultivate the use of digital pathology, including the computer-aided testing of samples, and to develop new AI tools that could be used to, for instance, match patients with optimal therapies.
Another aspect of the endeavor is to create a secure a data lake of thousands of curated images for building algorithms that can detect cancer automatically. A data lake is a storage repository that holds a vast amount of raw data in its native format.
Not only will the resulting data lake be made available to PathLAKE participants, but it will also be accessible to small- and medium-sized enterprises.
Snead said that despite advancements in digital pathology over the past decade, including the introduction of automatic slide scanners, such as Philips' Intellisite platform — which received US Food and Drug Administration clearance in April 2017 — there has been limited uptake of automation, and no AI to speak of. Instead, the UK's pathologists continue to personally assess each sample via a microscope slide, scoring them according to semi-quantitative methods.
"It's very much still an analog world at the moment," said Snead. He said that given the new funding, though, he expects the use of digital pathology to "take off in a big way," followed by the generation of deep learning methods, in part, because the scanning technology is there. "We now have the beginnings of producing digital pathology images at scale across the spectrum of human disease," Snead said. "This is all virgin territory for AI."
PathLAKE eventually aims to involve around 30 people, Snead said, two-thirds of whom will be computer science analysts and another third of whom will be pathologists. The resulting data lake of images will then be used in a number of envisioned projects involving biomarkers, automation, and to "develop products of interest" for the market.
"That will be the key thing to deliver, IP of commercial value," noted Snead. "It is the remit of the project to provide the UK with new IP that will generate income for the commercial sector."
Snead cautioned that it would be years before any new tools or algorithms related to identifying or grading cancers make it to the clinic. In addition to its activities in that area, Snead said that PathLAKE will be involved in education related to using AI and performing services for academic and commercial partners.
PathLAKE will also interact with the other four centers funded by Innovate UK, including iCAIRD in Scotland, which has a similar focus on applying AI in digital pathology and is also working with Philips. The Scottish center announced earlier this month that in addition to its £10 million investment from Innovate UK, a number of partner companies, including Philips, would invest an additional £5 million to support iCAIRD.
The new center will be headquartered at Queen Elizabeth University Hospital at the University of Glasgow. It also involves teams at the University of St. Andrews, the University of Edinburgh, the University of Aberdeen, and others.
David Harrison, the principal investigator on the project, is a professor of pathology at St. Andrews with visiting professorships at both Edinburgh and Glasgow. He said that the first aim of iCAIRD will be to convert the University of Glasgow's team of 47 senior pathologists to being fully digital over the next 18 months. Once that is accomplished, iCAIRD hopes to convert pathologists at the University of Edinburgh to being fully digital. Ultimately, roughly a million slides per year could be converted to images once iCAIRD succeeds in converting Scotland's pathologists.
"To have AI employed, we need digital use," noted Harrison.
As such, Harrison envisions that iCAIRD will employ around 10 staffers while it promulgates the use of digital pathology across Scotland. "One of our big drivers is to democratize access to quality diagnostics," noted Harrison. "It's not about centralizing," he said. "It's about ensuring however it's done, you will get the same high-quality standard of care."
While iCAIRD works to digitize Scottish pathology, it will also work with Philips to develop new algorithms. "We are not just beta testing what Philips already has," said Harrison. "We will be trying out new things with our AI team," he said. "This will be innovation."
AI need not ultimately produce a diagnosis either, Harrison stressed. If deep learning methods can reduce the amount of time spent per sample, then it will help lessen the burden for the UK's encumbered pathology labs. "If I can shave off three minutes each time a pathologist reads a slide, that's a lot of time saved," Harrison said.
In the longer term, Harrison said that iCAIRD will work to educate the next generation of pathologists to be familiar with digital pathology and AI. He speculated that at some point in the future, pathologists may become so comfortable with the technology that they might allow machines to make a diagnosis in certain cases. "I think it's more about informing the clinical decision," he cautioned. "I don't think there will be a bulk move to machine reading."
Peter Hamilton, group leader for digital computational pathology at Philips UKI, said that the company will provide the technology necessary to its new partners to fully digitize pathology laboratories.
Some of Philips' offerings that will be used include its Intellisite Pathology Suite, which includes slide-scanning technology and software for diagnostic image management; its TissueMark platform for automated detection of tumors in tissue samples; and "a range of tools to accelerate AI development in pathology," Hamilton said.
"Philips will pump-prime a number of centers, enabling them to fully demonstrate the benefits of digital transformation in this important discipline," Hamilton said. By increasing access to its technology, Philips aims to encourage the development of AI algorithms to support improved cancer diagnosis and treatment, he added.
Hamilton noted that the UK, by making its investment, is aspiring to "take a leading role as a nation in providing digitized pathology to patients." He said there is an increasing number of sites in Europe, as well as the US, that have moved completely to the digital reporting of slides. He likened the trend to the uptake of digital radiology, which he said is now embedded in most modern healthcare organizations.
"Digital pathology is such an important area of medical imaging for disease diagnostic and precision medicine, and it is only a matter of time before the whole discipline flips to digital workflows," said Hamilton. "The labs that have moved have shown efficiency gains of up to 25 percent — with clear benefits to turnaround times and the pathology shortages that the UK is currently facing," he said.
The adoption of AI, Hamilton added, should bring additional benefits.
In addition to PathLAKE and iCAIRD, a third center funded through the current investment that will focus on digital pathology is the Northern Pathology Imaging Cooperative (NPIC). The center announced earlier this month that it had received £10.1 million to expand a digital pathology and AI program across northern England. Investigators at the University of Leeds and Leeds Teaching Hospitals are leading the effort.
According to a statement, NPIC intends to place digital pathology scanners into a network of northern National Health Service hospitals, including all hospitals in West Yorkshire, the county surrounding Leeds, and Harrogate, a city to its north. This should generate about 760,000 images per year, about 1.2 petabytes of data annually that can be then used for training AI systems.
Leica Biosystems has been working with Leeds for years to digitize its pathology workflows. The company announced in December 2017 that pathologists at Leeds Teaching Hospitals NHS Trust and the University of Leeds had transitioned to full digital reads using Leica's automated scanning equipment, instead of using microscope slides. Last month, the company announced that the collaboration has so far resulted in "100 percent digital conversion."
Investigators at Leeds declined to provide additional details about the new center, noting that they had just received the funding. Leica meantime did not respond to emails seeking comment.