AI Fuels New Efforts in Computational Pathology

Labs can draw inspiration from Mayo Clinic on how to modernize clinical pathology workflows

CEO SUMMARY: Computational pathology combines technology and data science to improve laboratory medicine. Mayo Clinic is exploring how this new model can improve productivity and diagnostic accuracy in ways that even labs at smaller hospitals can put into practice. Success will stem from interdisciplinary cooperation between pathologists and data scientists.

MAYO CLINIC IS CURRENTLY SETTING THE STAGE for application of artificial intelligence (AI) and computational pathology to patient care within two years. 

Mayo Clinic’s new Division of Computational Pathology and AI is set to reimagine pathology and the laboratory through multidisciplinary teams of pathologists and data scientists. These teams will accomplish this by using technology, longitudinal information, and strategic partnerships.

The new division has a head start on this goal because it will build upon the Rochester, Minn.-based integrated health system’s earlier investments in digital pathology equipment, software, and vendor relationships. 

Clinical labs and anatomic pathology groups looking to take advantage of digital pathology will note that—while Mayo Clinic’s efforts benefit from the size of the institution—its approach to the new technology can be replicated by other organizations regardless of size.

“Computational pathology is the use of data science, information, and digital technologies for lab medicine,” stated Jason Hipp, MD, PhD, Chair of Computational Pathology and AI at Mayo Clinic’s Department of Laboratory Medicine and Pathology, and Director of Digital Innovation for Mayo Clinic Laboratories. He identified three factors that make these goals feasible: 

  • Cloud storage of abundant information;
  • Big datasets; and
  • AI, which has always needed large amounts of data and storage to be practical. 

Digital Opportunity for Labs

“We have a once-in-a-lifetime opportunity in this field,” he noted. “There are new opportunities to be generated with this new discipline—novel ways to generate revenue, conduct pathology, and rethink data.” Hipp, a pathologist and data scientist, spoke during a keynote at April’s Executive War College Conference on Laboratory and Pathology Management. The session was titled, “Building the Future of Computational Pathology and Artificial Intelligence at Mayo Clinic.”

Jason Hipp, MD, PhD, Chair of Computational Pathology and AI at Mayo Clinic’s Department of Laboratory Medicine and Pathology
Jason Hipp, MD, PhD

“Clinical laboratories need to redefine the practice of pathology digitally,” Hipp continued. “They must develop and integrate new tools to extract and analyze pathology and laboratory medicine data to bring new insights that do not exist right now. Doing so will bring value and operational efficiency to workflows and allow labs to offer more prognostic theory to clinicians.”

Artificial intelligence and machine learning will be transformative technologies in clinical labs and pathology practices. Forward-looking lab managers and pathologists will note that AI’s potential reaches beyond just large organizations such as Mayo Clinic. (See TDR, “Artificial Intelligence is Ready to Deliver for Labs,” July 26, 2021.)

For example, consider the following takeaways from AI experiences at Mayo Clinic, which even labs and pathology groups with limited budgets can explore:

Bring pathologists and data scientists together in multidisciplinary teams to learn from each other and collaborate to more effectively advance new initiatives in clinical settings.

Improve an organization’s use of digital image analysis beyond the medical laboratory by working in collaboration with radiology and other departments.

Computational Pathology?

Hipp observed that, amid a tight pathologist labor market and significant healthcare cost cutting, there is a need to enhance the productivity and responsibilities of individual pathologists. 

One approach to achieve this is to remove tedious tasks, like counting items, and elevate contributions to clinical assessment of patient cases. This is possible, Hipp said, through a combination of pathology, genomics, AI, and radiology. 

Computational pathology makes visible what is not apparent to the human eye, Hipp added. It is data driven, as compared to the subjective nature of traditional pathology. (See the sidebar below.) 

“Computational pathology enables digital analysis of every pixel in an image. These are smaller than cells,” he noted. “It enables a pathologist to see differences between cells and inflammatory cells. These are very novel features that cannot be quantified by pathologists working without these tools.” 

Boost Lab Workflows

Computational pathology advances digital pathology by integrating with machine learning for image analysis and precision medicine. Machine learning is a subset of AI. 

“We can take tissue, put it on a glass slide, image it, and make a diagnosis of cancer,” Hipp said. “Why don’t we insert an intermediary step? This is a step that would suggest medical patterns within the image that hold the keys to understanding therapeutic response to a prognosis. It is also the step we are working to achieve with the use of AI and related technologies.” 

Before arriving at Mayo Clinic, Hipp worked in pharma and technology companies. While at AstraZeneca, he and other colleagues wrote an article in 2020, titled, “The Revival of the H&E with Artificial Intelligence” for the Journal of Clinical and Anatomic Pathology. 

The article spotlighted the following ways that computational pathology and AI can aid diagnostics workflow:

  • Directing the pathologist to a region of the slide.
  • Determining cancer presence or absence.
  • Providing tumor grading.
  • Searching for and identifying diagnostically similar cases.
  • Predicting mutational status from hematoxylin and eosin (H&E) stains.

“I consider computational pathology to be a new frontier for our profession,” Hipp observed. “I think of this as a revival of the H&E, which is so vital to pathology.”

As The Dark Report previously noted, Mayo Clinic is bridging the work of pathologists and data science engineers who develop AI algorithms. (See TDR, “Innovation Showcased at Executive War College,” May 16, 2022.)

“We want to embed both sets of skills into the pathology workflow,” Hipp explained. “We need the data scientists working side by side with the pathologists. That practical part is missing today.” Hipp experienced a similar workflow when he worked previously as a pathologist at Google Health alongside engineers. 

Hipp explained that pathologists look at things differently than data scientists, so it benefits computational pathology projects to have both disciplines collaborating. “There’s a lot of discovery that requires data scientists and pathologists to work very closely together,” Hipp noted.

Digitizing Millions of Slides

One ambitious project now unfolding within three years at Mayo Clinic’s Department of Laboratory Medicine and Pathology is the goal of digitizing five million of its 25 million archived tissue glass slides. This digital data—stored in a cloud-based image repository—will be accessible for use in clinical pathology, as well as for research and education. 

“We don’t want to create data silos,” Hipp declared. “Rather, we look at this as a ‘data tumor board,’ where all the data is available for evaluation. Because it’s integrated, it’s expected to produce novel insights and better patient care.” 

There is another rich source of data that Hipp’s team wants to incorporate into the diagnostic/therapeutic process. Currently, Mayo Clinic stores three billion paper pathology reports going back more than 100 years. They, too, will be digitized to support the pathology slides. 

Longitudinal Information 

Such longitudinal information and history of patient care has value to the creation of algorithms. “What we want to do with all this longitudinal data—pathology data, radiology data, genomics data—is determine what information we can identify for algorithms,” Hipp said.

Integrating algorithms with the data will be a critical part of the work ahead. The efforts fall into three buckets: 

  • Detection and triage. 
  • Quantification.
  • Prediction.

“The algorithmic tools will be leveraged for enlightenment of detection, but more so for workflow decision-making,” Hipp said. “What is most exciting to me is predictive value: How do we know whether a patient will respond to an immmunotherapy drug? And what is the risk and reward?” 

It’s not just pathology that will be a primary focus for Mayo Clinic’s Division of Computational Pathology and AI. The division aims to advance medical imaging throughout the institution starting with Mayo Clinic Laboratories, the Department of Laboratory Medicine, the Mayo Clinic Platform, and Biopharma. 

“We are going to first empower these divisions to leverage this technology and then work in other layers of radiology, digital health, and informatics,” Hipp said. “We are trying to implement a matrix approach that is currently not being done in traditional patient care settings.” 

Hospital-based pathologists should consider this broad approach to digital and computational pathology and reach out to peers in other areas of the organization for assistance, including data scientists.

Contact Jason Hipp, MD, PhD, at

Step-by-Step Plan for Computational Pathology Guides Team at Mayo Clinic Laboratories

WITHIN THE NEW DIVISION OF COMPUTATIONAL PATHOLOGY AND ARTIFICIAL INTELLIGENCE AT THE MAYO CLINIC, the goal is to pull together a host of technologies and integrate them to boost the productivity of pathologists while also improving the accuracy of diagnostic and therapeutic services. During his presentation at last April’s Executive War College, Jason Hipp, MD, PhD, Chair of Computational Pathology and AI at Mayo Clinic’s Department of Laboratory Medicine and Pathology, and Director of Digital Innovation for Mayo Clinic Laboratories, described how computational pathology is expected to change the workflow for anatomic pathologists. 

Benefits of Computational Pathology and AI

  • To aid the pathologist in the diagnostic workflow
  • Directing the pathologist to a region of the slide
  • Determine the presence/absence of cancer
  • Provide the tumor’s grading
  • To quantify biomarker (IHC) expression
  • To search and identify diagnostically similar cases
  • To predict mutational status from H&E
  • To predict protein (IHC) expression from H&E
  • To predict risk category as determined by gene expression from H&E
  • To predict survival

SOURCE: The revival of the H&E with AI; Burlutskiy et al. Journal of Clinical & Anatomic Pathology (2020)

Traditional Pathology vs Computational Pathology

Traditional pathology is limited and subjective

  • Tissues and cells are examined under a microscope 
  • Subjective, qualitative, and semi-quantitative assessments are made about the tissue
  • Pathology report contains both qualitative and semi-quantitative interpretations
  • Human brain looks for specific patterns and ignores non-confirmatory data


Significant amount of data from tissue remains unutilized or underutilized


Computational pathology is robust and data driven

  • Tissues and cells are examined by computer algorithms
  • Robust, highly quantitative and complex spatial measurements generated for every pixel/feature
  • Numerical results are integrated with disparate datasets and analyzed by secondary analytic tools
  • Computer brain ‘looks’ at everything and can identify novel features


Most data from tissue is utilized


External Partners Aid in Pathology Transformation 

MAYO CLINIC IS WORKING WITH VENDORS to digitize and analyze its clinical pathology data.

Pramana, an AI company in Cambridge, Mass., that focuses on the pathology sector, is helping Mayo Clinic digitize an initial five million glass pathology slides. Pramana sells a whole-slide imaging system—fed by a robot and analyzed by AI algorithms—to make automated quality assessment of slides possible. Technology from Pramana can scan more than 1,000 slides per day, according to the company.

Meanwhile, as part of a partnership between Google and Mayo Clinic, Google Cloud is storing the health system’s data and applying AI to it. Google, headquartered in Mountain View, Calif., opened an office in Rochester, Minn., in 2021 so that its engineers can work closely with Mayo Clinic researchers. 



You are reading premium content from The Dark Report, your primary resource for running an efficient and profitable laboratory.

Get Unlimited Access to The Dark Report absolutely FREE!

You have read 0 of 1 of your complimentary articles this month

Privacy Policy: We will never share your personal information.