Exploring the Benefits of Digital Pathology Platforms and Sharing

December 18, 2023 - Last month, at the Radiological Society of North America’s (RNSA) Scientific Assembly and Annual Meeting, Paige announced its partnership with Nuance, a Microsoft company, to develop a digital pathology and consultation network for pathologists. The collaboration aims to provide a secure and accurate data-sharing network that streamlines and improves the diagnostic pathway in oncology. The companies are exploring the benefits and applications of digital pathology in the evolving healthcare environment.

The Collaboration

According to the RNSA annual meeting announcement, the collaboration between Paige and Nuance combines Paige’s digital pathology tools, including FullFocus and FullFolio, with Nuance’s image-sharing network, PowerShare.

“It enables people to read slides using a monitor instead of using that traditional microscope,” explained Razik Yousfi, senior vice president of technology for Paige, in his interview with LifeSciencesIntelligence. “Paige enables institutions — when they receive the scanner — to start looking at those images digitally instead of the glass itself.”

He clarified that the current gold standard for pathology is using a microscope.

The collaboration with Nuance provides access to its application, PowerShare. Yousfi noted that PowerShare is a pathology-sharing network available in over 14,000 sites across the United States that allows secure sharing and viewing of radiological imaging.

By combining PowerShare with Paige’s existing digital pathology technology, the companies can offer a digital data-sharing platform for radiologists and pathologists.

“The platform at Paige is already somewhat unified, but we don't have the footprint that enabled us to share images within a super large network,” noted Yousfi. “Nuance, on the other hand, has the technology and product that covers a large network in the radiology space and not in the pathology space.”

Benefits of Digital Pathology

Considering the conventional gold standard, LifeSciencesIntelligence set out to understand the benefits of a digital platform compared to traditional microscopy.

Yousfi explained that, at the most basic level, shifting to a digital platform widens access to the tools used to analyze radiological images. For example, finding a suspicious area in an image, measuring it, and reporting the details is traditionally done using a ruler under a microscope. However, going digital provides a set of tools that simplify the process.

“Some of the tedious tasks associated with reading glass slides can be simplified by doing them on a monitor, and because [health systems are] going digital, [providers] have access to a set of capabilities that you don't have otherwise,” he continued.

Beyond quickly measuring imaging anomalies, providers using this platform can easily share images with colleagues to get another opinion from pathologists in the same institution.

In addition to sharing the basic image, pathologists can provide all the context and information digitally, minimizing the tedious tasks of in-person slide analysis.

Another advantage of digital scans and sharing technology is the ability to tag or mark slides for future reference. Using that tool, providers can select images to incorporate into education, research, or other practices.

“There are a bunch of advantages associated with going digital. There are a set of publications that have been done independently off Paige and with Paige that show that there is, traditionally, a boost in productivity associated with going digital,” revealed Yousfi.

AI Integration

Besides procedural advantages, Yousfi highlighted that going digital allows health systems to leverage artificial intelligence (AI).

“Going digital is the springboard to use a much more sophisticated set of applications and AI to help with some of the more challenging cases and tedious tasks,” added Yousfi. “At Paige, on top of the platform, we have been developing a set of artificial intelligence solutions.”

For example, Paige has FDA approval for a tool that assists pathologists in detecting cancer in prostate needle biopsies. By applying digital imaging, Paige can integrate that tool into the pathology platform, allowing providers to use AI technology to ensure they have not missed a cancer.

“Without going digital, none of those AI applications are accessible to providers,” Yousfi stressed. “Generally speaking, going digital enables [health systems] to adopt AI solutions, and those AI solutions help with an increase in the confidence at which [providers] make their diagnosis.”

Patient Care

After touching on procedural benefits and technology advantages, Yousfi explained to LifeSciencesIntelligence that digital pathology and platforms combining digital pathology and secure patient sharing offer unparalleled patient benefits.

“It's important to note that pathology is specialized,” he began. Various pathologists specialize in different types of cancer. That means it is common for pathologists to send slides for a second opinion or consultation, especially if the cancer in question is not their specialty. Most of the pathology volume at larger academic medical centers consists of second opinions.

“In other words, glass slides get prepared in a particular location, and a pathologist looks at them. They may not necessarily want to commit to the final diagnosis, and they may send that to another pathologist specializing in that academic medical center to provide them with their own view on the diagnosis,” Yousfi explained.

With the current analog standard, providers who do not work at these larger institutions prep the glass slides and ship them to the appropriate academic medical center, where they are input into the system. At that point, the consulting pathologist analyzes the slides, generates the report, and sends back the information, including the report and slides, to the originating institution.

While that process can provide valuable insights, it is complex and, at times, inefficient. Yousfi explained that the shipping process can be lengthy and costly, burdening patients. For example, patients who live in rural areas, far from these larger institutions, often must engage in these procedures. Yousfi explained that, often, smaller local community hospitals are unable to diagnose rarer cancers, so they send the slides out this way.

One of the challenges is that patients may end up with the wrong diagnosis, depending on the potential complications that slides may experience in transit or miscommunications between providers.

Beyond that, the wait period between a consult and a diagnosis can be up to six weeks, meaning patients could spend over a month awaiting a second opinion. During that time, patients are unaware if they have cancer, its severity, or the potential treatment options. From a patient's perspective, these delays can be nerve-wracking.

However, the process can shrink to just a few hours with digital pathology. As a result, patients have timely responses and can initiate treatments earlier with insights from a specialized pathologist.

“Ultimately, that's the main reason why we do this. It's to enable patients to access better, faster care.”

Minimizing AI Bias

Although integrating AI can provide valuable insights into patient health, models must be trained and developed in a way that minimizes bias. LifeSciencesIntelligence asked how Paige intended to mitigate AI bias.

“Not every AI is built, trained, and deployed the same way,” began Yousfi.

To combat AI bias, Paige is deploying multiple procedures. According to Yousfi, most AI systems are trained on a few slides pathologists manually annotate. However, he points out two major issues with that process. First, that process is difficult to deploy at a larger scale because it requires access to many pathologists. Additionally, the conclusions made by pathologists may be incongruent with one another. For example, they may not agree on cancer severity, which leads to poorly curated data that the system is based on.

“Back in 2018 and 2019, we published our approach to a new way of building AI systems that leverage multiple instance learning in Nature Medicine,” revealed Yousfi. 

This approach does not use manual annotations. Instead, it incorporates large data sets that include images and their corresponding clinical reports. The AI system finds correlations between clinical reports and images instead of using manual annotations. There are two benefits to this methodology: (1) more data input yields better results, and (2) there is minimal bias because the model does not learn from manual annotations.

“The second thing is data source and data provenance. A lot of the time, institutions build AI with slides that have been maybe digitized in one location, and that's a problem,” Yousfi added.

A singular data source can cause multiple conflicts and contribute to bias. For example, the slides may be prepared in a certain way or at a center with a uniform patient population. If the AI is trained on these slides, it may result in some bias.

“When it comes to the datasets we're using, we have access to images that are being digitized from glass slides coming from more than 800 institutions and more than 45 countries, so our data sets for AI development have been diverse from the get-go,” explained Yousfi. “It's a real-world data set, and because we've used it, we were able to generalize so well that the system can be deployed without any calibration to any institution.”

Digital Pathology Integration

Although he admits that digital integration takes time, he maintains that productivity improves significantly with adopting digital tools such as this one.

As Yousfi mentioned, getting a health system to integrate a digital pathology platform can take some time. He noted that the integration process requires two basic materials: the scanner to digitize slides and the actual platform.

Although Paige does not provide or build scanners, their role begins right after digitization.

Instead of manufacturing the scanners, Paige has partnered with scanner manufacturers, including Leica Biosystems, to allow institutions to purchase and install functional scanners.

“That helps provide this end-to-end turnkey solution. We've also announced this end-to-end solution where very recently, if [healthcare facilities] get a GT 450 scanner, [they] have everything integrated and, as an institution, can go digital faster.”

Once a scanner is available, Paige allows institutions to integrate their laboratory information systems in the Paige ecosystem and store digitized images of glass slides at an appropriate scale that represents the imaging volumes of that institution.

Beyond that, the platform provides them with the entire digital ecosystem, including tools for slide management, viewing, and AI applications. Paige also offers assistance in training providers and health systems.

“We also collaborate with organizations like Microsoft to really do that in a secure, state-of-the-art way, where information is managed in the most sensitive way,” added Yousfi.

The system complies with the Health Insurance Portability and Accountability Act, a policy that enforces health data security in the US, and the General Data Protection Regulation, a parallel requirement in the European Union. Additionally, the Paige platform has been ISO 2700 certified, meaning it follows a set of information security standards developed by the International Organization for Standardization (ISO) and the International Electrotechnical Commission (IEC).

“Through the relationship with Nuance, we can very rapidly expand the set of institutions that can receive images, and this is where the partnership shines,” stated Yousfi.

The platform for Paige has already been launched, and the Nuance network has been around for a long time, but the integration was recently announced. Moving forward, the companies hope to recruit health systems and deploy the product.

“We're actively discussing with early customers and partners to have them participate in this collaboration. We will announce them very shortly, but we also plan on global, generally announcing this product early next year. That's imminent; we're talking Q1,” Yousfi revealed.

Editor's Note: This article has been edited to correct digital radiology to digital pathology based on clarifications from Paige.