banner of LMP department's DP&AI initiative

Informatics spurs pathology’s digital transformation

|
William Hoffman
4 minutes
Patients will benefit as pathology enters a new era

Digital pathology and artificial intelligence imageSixty years ago, the University of Minnesota’s College of Medical Sciences established its first computer facility. The new Biomedical Data Processing Unit partnered with Minneapolis-based Control Data Corporation (CDC), then a computing powerhouse. CDC’s 3000 series mainframe computers designed by supercomputer pioneer Seymour Cray, a University graduate, gave the institution an early advantage in health informatics. Historian Dominique A. Tobbell called it “a defining moment in the early history of health informatics nationally” in her book Health Informatics at Minnesota: The First Fifty Years. Laboratory Medicine and Pathology was instrumental in establishing the University’s Institute for Health Informatics, which is celebrating its 50-year anniversary.

Early computing supported advances in graphical displays, statistical classification, medical diagnosis, data storage and retrieval, simulation, signal processing, and laboratory information systems. The foundational concepts and algorithms of the 1950s and 60s underpin today’s machine learning, artificial intelligence (AI), and digital health applications. Modern informatics continues that trajectory, transforming raw data into actionable insights to improve patient care.

Michell Stoffel photo

Michelle Stoffel, MD, PhD, LMP assistant professor of pathology, director of clinical informatics, and associate chief medical information officer for M Health Fairview, works at the center of this transformation. Her role focuses on integrating laboratory results from the laboratory information system (LIS) into patients’ electronic health records (EHRs). In digital pathology, this means embedding whole slide images, often gigapixel (a billion pixels) in scale, directly into the EHR.

“As the medical informatics director, I see myself as a bridge between laboratories and the broader digital health ecosystem,” Stoffel explained. “At M Health Fairview, we have a hybrid model: we routinely scan and interpret slides, and those whole slide images are accessible within the EHR. It’s not yet seamless, but it represents real progress.”

Currently, integration relies on third-party software such as Leica’s Aperio eSlide Manager, accessed via hyperlinks within Epic Beaker. This ensures images are correctly linked to patient records, improving safety, but the workflow is limited and lacks full bidirectional data flow. “It’s a big improvement over not having integration,” Stoffel said. “But right now, it’s still just a hyperlink, not a full-interoperability system.”

Looking ahead, Stoffel envisions a system where annotations, AI analyses, and EHR data can flow seamlessly between the image management system and the medical record. “That’s the future state,” she said. “But today, we’re in the early phases, with limited adoption and functionality.” Achieving this will require technical investment and careful oversight, given the regulatory environment. 

Validation remains a central challenge. “The high accreditation standards in pathology create an environment of safety. But this safety comes with a resource cost.  We have to validate everything,” Stoffel emphasized. “For CAP and CLIA compliance, every change requires documentation and proof. Even seemingly minor updates can trigger an extensive revalidation process.”

Stoffel sees digital pathology paralleling earlier advances in clinical chemistry, where standardization and quality control paved the way for widespread adoption. “The biggest lift is often upstream, such as standardizing histology processes before slides are even scanned,” she said. “Institutions that succeed in digital pathology invest heavily in quality and consistency.”

Despite these hurdles, Stoffel believes M Health Fairview is well-positioned for scale-up. “We’re ahead of most institutions,” she noted. “Only a handful of academic centers nationwide are doing remote digital sign-out. We’ve started one tier of services and are positioned to build on that in the future.”

Digital pathology will also reshape how future pathologists are trained. “For residents, whole slide imaging changes everything about accessibility,” Stoffel said. “Instead of being limited to a glass slide that only one person can look at under a microscope, digital images can be shared, saved, and multiply their educational impact. Digital pathology will facilitate collaborative learning, make it easier to access rare cases, and integrate with AI tools that will soon be the new foundational tools on which practice depends and which trainees need to learn how to use and manage. It’s going to revolutionize how trainees learn and how we teach.”

Stoffel believes practicing pathologists will also benefit greatly from the coming practice changes. “In the future, digital pathology will allow us to work more flexibly, collaborate across institutions and even continents, and standardize practice to the degree needed to meaningfully implement AI,” she said. “This isn’t about replacing pathologists. I don’t think that will be happening in the foreseeable future. Rather, it’s about giving us better tools to make the most of the workforce and knowledge that we have.”

Here are the stories in our DP & AI series: