Recognizing the increasing importance of digital pathology and its potential impact to transform federal health care, government, military, and university digital pathology specialists convened in May 2023 to share expertise to advance the use of digital pathology in federal health care.
The seminar was hosted by the University of Pittsburgh and led by Ronald Poropatich, MD, Director of the Center for Military Medicine Research, Health Sciences, and Professor of Medicine at the University of Pittsburgh Medical Center, and Douglas Hartman, MD, Vice Chair of Pathology Informatics, Associate Director of the Center for AI Innovation in Medical Imaging, and Professor of Pathology at the University of Pittsburgh/University of Pittsburgh Medical Center (UPMC).
Invitees included senior federal government pathologists, laboratory scientists, IT leaders, and stakeholders from the VA, DoD, HHS (NIH, CDC, IHS, FDA) and other federal agencies. The speakers for the conference were CDR Roger Boodoo, MD, Chief of Innovation, Defense Health Agency; Ryan Collins, MD, Pathologist, Williamsport Pathology Association; Pat Flanders, Chief Information Officer, J6, Defense Health Agency; Matthew Hanna, MD, Director, Digital Pathology Informatics, Memorial Sloan Kettering Cancer Center; Stephanie Harmon, PhD, Staff Scientist, NIH NCI, Imaging/Data Scientist in Molecular Imaging; Douglas Hartman, MD, Vice Chair of Pathology Informatics, University of Pittsburgh; Stephen Hewitt, MD, PhD, Head, Experimental Pathology Laboratory, NIH NCI, Center for Cancer Research; Jason Hipp, MD, PhD, Chief Digital Innovation Officer, Mayo Collaborate Services, Mayo Clinic; Brian Lein, MD, Assistant Director, Healthcare Administration, Defense Health Agency; Col Mark Lyman, MD, Pathology Consultant to the US Air Force Surgeon General; COL Joel Moncur, MD, Director, Joint Pathology Center; Ronald Poropatich, MD, Director of the Center for Military Medicine Research, Health Sciences; Professor of Medicine, University of Pittsburgh; David Shulkin, MD, Ninth U.S. Secretary of Veterans Affairs; Eliot Siegel, MD, Chief of Radiology and Nuclear Medicine, Veterans Affairs Maryland Healthcare System; Professor and Vice Chair, University of Maryland School of Medicine; CDR Jenny Smith, DO, Pathologist, US Naval Medical Center Portsmouth; Shandong Wu, PhD, Associate Professor, Departments of Radiology, Biomedical Informatics, and Bioengineering, Director of Center for Artificial Intelligence Innovation in Medical Imaging, University of Pittsburgh; LCDR Victoria Mahar MD, Pathologist, US Army.
Throughout the 1.5-day meeting, topics such as the integration of systems, the value of single vendor solutions vs multiple vendors, and the interconnectedness of radiology and pathology in health care were discussed. The speakers addressed the challenges of adopting digital pathology, including workflow improvement, quality control, and the generalizability of algorithms. The importance of collaboration, leadership, data analytics, compliance with clinical practice guidelines, and research and development efforts were stressed. The increasingly important role of artificial intelligence (AI) in digital pathology, its applications, and its benefits were also highlighted. Continuing education credits were offered to participants.
Overall, the meeting provided valuable insights into the advancements, challenges, and potential of digital pathology, AI, and technology integration in the federal health care ecosystem. However, this cannot be achieved without leadership from and close collaboration between key industry, academic, and government stakeholders.
Uses of Digital Pathology
Digital pathology refers to the practice of digitizing glass slides containing tissue samples and using digital imaging technology to analyze and interpret them. It involves capturing high-resolution images of microscopic slides and storing them in a digital format. These digitized images can be accessed and analyzed using computer-based tools and software.
While traditional pathology involves examining tissue samples under a microscope to make diagnoses and provide insights into diseases and conditions, digital pathology uses digital scanners that capture all relevant tissue on the glass slide at high magnification. This process generates a high-fidelity digital representation of the tissue sample that can be navigated akin to how glass slides are reviewed on a brightfield microscope in current practice (eg, panning, zooming, etc). Microscopic review of patient specimens in pathology allows for identifying patterns and markers that may not be easily detectable with manual examination alone.
The digitized slides can be stored in a database or a slide management system, allowing pathologists and other healthcare professionals to access and review them remotely, thus creating the potential to improve collaboration among pathologists, facilitate second opinions, and enable easier access to archived slides for research purposes.
Potential Benefits
Digital pathology also opens the door to advanced image analysis techniques, such as computer-aided diagnosis, machine learning, and AI algorithms, with the potential for the following outcomes and benefits:
- Improved accuracy AI algorithms can analyze large volumes of digital pathology data with great precision, reducing the chances of human error and subjective interpretation. This can lead to more accurate and consistent diagnoses, especially in challenging cases where subtle patterns or features may be difficult to detect.
- Automated detection and classification AI algorithms can be trained to detect and classify specific features or abnormalities in digital pathology images. For example, AI models can identify cancerous cells, tissue patterns associated with different diseases, or specific biomarkers. This can assist pathologists in diagnosing diseases more accurately and efficiently.
- Quantitative analysis AI can analyze large quantities of digital pathology data and extract quantitative measurements. For instance, it can calculate the percentage of tumor cells in a sample, assess the density of immune cells, or measure the extent of tissue damage. These objective measurements can aid in prognosis prediction and treatment planning.
- Image segmentation AI algorithms can segment digital pathology images into different regions or structures, such as nuclei, cytoplasm, or blood vessels. This segmentation allows for precise analysis and extraction of features for further study. It can also facilitate the identification of specific cell types or tissue components.
- Image enhancement AI techniques can enhance the quality of digital pathology images by improving clarity and reducing noise or artifacts. This can help pathologists visualize and interpret slides more effectively, especially in challenging cases with low-quality or complex images.
- Decision support systems AI-powered decision support systems can assist pathologists by providing recommendations or second opinions based on the analysis of digital pathology data. These systems can offer insights, suggest potential diagnoses, or provide relevant research references, augmenting the pathologist’s expertise and improving diagnostic accuracy.
- Collaboration and second opinions Digital pathology, combined with AI, facilitates remote access to digitized slides, enabling pathologists to seek second opinions or collaborate with experts from around the world. This can enhance the quality of diagnoses by leveraging the collective expertise of pathologists and fostering knowledge sharing.
- Education and training AI algorithms can be utilized in virtual microscopy platforms to create interactive and educational experiences. Pathology residents and students can learn from annotated cases, receive real-time feedback, and develop their skills in a digital environment.
- Research and discovery AI can assist in identifying patterns, correlations, and novel biomarkers in digital pathology data. By analyzing large datasets, AI algorithms can help uncover new insights, contribute to research advancements, and aid in the development of personalized medicine approaches.
- Predictive modeling AI can analyze vast amounts of digital pathology data, patient records, and outcomes to develop predictive models. These models can estimate disease progression, treatment response, or patient survival rates based on various factors. They can contribute to personalized medicine by assisting in treatment decisions and prognosis assessment.
It is important to note that while AI has shown promising results, it is not intended to replace human pathologists but to augment their capabilities. Overall, the combination of AI technology with the expertise of pathologists can lead to improved diagnosis, better patient care, and more efficient workflows in digital pathology.