The Evolving Role of Pathologists
The future of AI in pathology goes beyond image analysis alone. Increasingly, AI systems are integrating diverse data types to provide richer diagnostic, prognostic, and therapeutic insights. This module explores the rise of multimodal AI, the importance of building robust data ecosystems, and how pathologists are positioned to lead the AI-augmented transformation in healthcare.
Beyond Images: The Rise of Multimodal AI in Pathology
Traditional pathology focuses on morphological assessment of tissue slides. However, new AI models are combining pathology images with transcriptomics, radiologic imaging, and electronic health record data to create comprehensive disease profiles. For example, in breast cancer, integrating histopathology with gene expression and radiology features has improved subtype classification and therapy response prediction (Chen et al., 2025). In gliomas, combining MRI, histology, and molecular data enables more accurate grading and prognosis (Wang et al., 2024). Multimodal approaches also facilitate diagnosis of rare diseases where any single data source may be insufficient.
These AI tools demand careful validation that reflects their complex inputs and intended clinical applications. Multimodal validation requires multi-institutional datasets that encompass the diverse data types involved and close collaboration between pathologists, radiologists, and clinicians (Liu et al., 2023).
Building a Data Ecosystem for AI-Enabled Pathology
Data is the foundation for any successful AI deployment. Adhering to FAIR principles (Findable, Accessible, Interoperable, and Reusable) ensures that slide images and metadata can be effectively leveraged for training and inference (Wilkinson et al., 2016). Structuring slide archives alongside detailed metadata, including staining protocols, scanner settings, and patient demographics, is crucial.
Annotation practices vary, but consensus labeling is widely regarded as best practice to reduce bias and increase dataset reliability. While solo labeling may speed up initial dataset creation, consensus approaches involving multiple expert pathologists increase the quality and reproducibility of annotations (Ourselin et al., 2022).
Partnerships with AI vendors require clear agreements around data governance and intellectual property. Collaboration with academic groups often emphasizes open datasets and reproducibility, offering complementary benefits. Balancing these approaches enables pathology departments to cultivate a sustainable AI data ecosystem.
The Pathologist’s Role in the AI Era
AI is not poised to replace pathologists but to augment their expertise and extend their impact. The emerging clinical paradigm centers on human-AI collaboration, with pathologists interpreting AI outputs and integrating multimodal data into nuanced diagnoses.
Leading digital transformation within hospital systems will increasingly fall to pathologists equipped with AI literacy, including an understanding of model limitations, validation strategies, and governance responsibilities. Upskilling in these areas is essential to maintain clinical authority and advocate for ethical AI use.
This new role also encompasses educating colleagues and trainees on AI capabilities and pitfalls, participating in multidisciplinary teams, and driving innovation to improve patient outcomes (Topol, 2019).
AI Evaluation and Case Studies
To support pathologists in navigating the complex AI landscape, standardized evaluation tools are invaluable. One such resource is the AI Evaluation Checklist developed by the Royal College of Pathologists and other professional bodies. This framework guides users through critical aspects such as data provenance, algorithm transparency, clinical validation, integration readiness, and post-deployment monitoring (RCP, 2024). Downloadable and adaptable, it serves as a practical aid to systematically assess AI tools before clinical adoption.
Real-world implementations offer rich lessons. Academic centers like the Mayo Clinic have documented their multi-year journey selecting and deploying AI algorithms, emphasizing early involvement of pathologists in vendor discussions and workflow redesign (Aiforia, 2024). Similarly, Radboud University Medical Center’s experience highlights the importance of rigorous validation with local data and iterative feedback loops with end users to improve AI acceptance (van der Laak et al., 2023). At MD Anderson Cancer Center, integration of AI into routine workflows was coupled with structured training programs, fostering clinician confidence and sustained use (MD Anderson, 2023).
These case studies converge on common themes: the necessity of transparent vendor engagement, robust validation in local context, and proactive physician education. Addressing these elements will accelerate successful AI adoption and maximize clinical impact.
References
Royal College of Pathologists (RCP). (2024). Artificial Intelligence in Pathology: A Guide for Pathologists. rcpath.org
Aiforia. (2024). How Mayo Clinic Chose a Vendor for AI in Pathology. aiforia.com
van der Laak, J., et al. (2023). Real-world validation of AI tools in pathology. Nature Medicine.
MD Anderson Cancer Center. (2023). Artificial Intelligence in Pathology. mdanderson.org
