What can AI pathology models actually do today?
AI pathology models are already moving from research labs into real-world clinical workflows, particularly in cancer diagnosis. A 2024 review found that several FDA-approved digital scanners and image management systems are establishing a solid foundation for integrating AI into everyday pathology practice, potentially reducing device and operational costs in the future [1]. The same review notes that AI is transforming how morphological diagnoses are automated, notably in cancers like prostate and colorectal, within screening initiatives [1].
More advanced AI models, called foundation models, are also emerging. A 2023 review explains that these large AI models, trained on vast datasets, can be fine-tuned for specific tasks like quantifying image analysis, generating pathology reports, and even predicting prognosis [3]. A 2025 study adds that these models have shown remarkable capabilities in disease diagnosis, rare cancer detection, and biomarker expression prediction [5]. However, the 2025 study also cautions that clinical application faces challenges like data quality, interpretability, and scalability [5].
What's holding AI pathology back from widespread use?
Despite the promise, there's a significant gap between what AI can do in controlled studies and what actually gets adopted in clinics. A 2022 paper on designing AI-augmented healthcare systems identifies four critical pillars for adoption: physician buy-in, patient acceptance, provider investment, and payer support [6]. Without all four, even technically excellent AI tools may sit on the shelf. The paper emphasizes that how physicians integrate AI into their practice and how patients perceive AI's role are key barriers [6].
Ethical concerns also loom large. A 2021 review on AI ethics in pathology warns that if used unethically, AI may exacerbate existing health care inequities, especially if not implemented correctly [4]. It highlights three foundational principles—transparency, accountability, and governance—that must guide future practice [4]. Additionally, a 2024 survey of medical students found that while 91% felt ready for AI in healthcare, their actual awareness of AI scored only 45%, and confidence scored 58% [2]. This gap between enthusiasm and knowledge suggests that education and training are lagging behind technological advances [2].
What needs to happen for AI pathology to reshape healthcare delivery?
For AI pathology to truly reshape healthcare delivery over the next decade, several pieces must fall into place. First, regulatory frameworks need to evolve. A 2024 review notes that the U.S. regulatory environment, shaped by the FDA, CMS/CLIA, and CAP, is adapting to accommodate AI innovations while ensuring safety and reliability [1]. New digital pathology CPT codes are also facilitating reimbursement, which is critical for provider investment [1].
Second, the technology itself must overcome current limitations. A 2023 review points out that task-specific AI models often struggle to generalize to new datasets or unseen variations in image acquisition, staining techniques, or tissue types [3]. Foundation models offer a path forward by providing in-context learning and the ability to self-correct, but they require massive amounts of data and careful validation [3][5].
Finally, the human element is crucial. A 2022 paper argues that AI-augmented systems must be designed with purposeful attention to physician workflows and patient trust [6]. The 2021 ethics review insists that pathologists must have a seat at the table to drive future implementation of ethical AI [4]. Without addressing these social and organizational factors, even the best AI models will fail to deliver on their promise.
About These Sources
This answer is built on 6 peer-reviewed studies — published from 2021 to 2025, 3 from 2024 or later, 4 in Q1 journals, collectively cited 294 times — selected as the most relevant from 7 studies that passed quality screening, drawn from 69 papers retrieved from a database of over 500 million.
Sources used in this answer
Implementation of Digital Pathology and Artificial Intelligence in Routine Pathology Practice
A 2024 review found that FDA-approved digital scanners and image management systems are creating a foundation for AI integration in pathology, with AI automating diagnoses in prostate and colorectal cancer screening, though challenges like data privacy and algorithmic bias remain.
Understanding AI in Healthcare: Perspectives of Future Healthcare Professionals
A 2024 cross-sectional survey of 217 medical students in India found that while 91% felt ready for AI in healthcare, actual AI awareness scored only 45%, and confidence scored 58%, highlighting a gap between enthusiasm and knowledge.
Revolutionizing Digital Pathology With the Power of Generative Artificial Intelligence and Foundation Models
A 2023 review explains that foundation models and generative AI can be fine-tuned for tasks like quantifying image analysis, generating pathology reports, and prognosis, overcoming limitations of task-specific models that struggle to generalize to new data.
Ethics of AI in Pathology
A 2021 review on AI ethics in pathology warns that AI may exacerbate healthcare inequities if used unethically, and emphasizes three foundational principles—transparency, accountability, and governance—that must guide future practice.
Foundation Models in Digital Pathology Imaging: Next-Generation AI for Healthcare Transformation
A 2025 study on foundation models in digital pathology found they show remarkable capabilities in disease diagnosis, rare cancer detection, and biomarker prediction, but face challenges with data quality, interpretability, scalability, and regulatory compliance.
Designing AI‐augmented healthcare delivery systems for physician buy‐in and patient acceptance
A 2022 paper on AI-augmented healthcare systems identifies four pillars for adoption—physician buy-in, patient acceptance, provider investment, and payer support—and stresses that design must account for how physicians integrate AI and how patients perceive its role.