How AI is Advancing Cancer Diagnosis

The following is a guest article by Razik Yousfi, CEO and CTO at Paige

As AI gains traction in healthcare, it is demonstrating enormous potential to transform cancer diagnostics and serve as a powerful partner to pathologists—many of whom have been using glass slides and microscopes for over 150 years. We sat down with Paige CEO, Razik Yousfi, to learn more. Paige is recognized for its industry-leading AI technology, regulatory approvals, and development of the largest foundation models in pathology, built upon one of the world’s most comprehensive databases originating from a world-renowned cancer center. As one of the first to bring novel AI technology and integrated diagnostics into clinical pathology, Paige is shaping the future of cancer detection.

Paige pioneered AI-driven pathology solutions, receiving the first-ever FDA approval for its AI-powered cancer detection tool in 2021, and has since expanded its suite of models to detect cancer across 40+ organs and tissue types, including common, rare, and precursor lesions. Its partnerships with leading commercial labs, digital pathology providers, and healthcare institutions are accelerating AI adoption worldwide.

Razik brings deep technical expertise and a strategic vision for scaling AI adoption in clinical practice and using foundation model technology to help research institutions and pharmaceutical companies navigate the complexities of cancer. In this interview, he shares his insights on the evolving role of AI in cancer diagnostics and more. 

What are the Most Promising Applications of AI in Improving Patient Outcomes and Streamlining Healthcare Operations?

AI is enhancing cancer diagnostics by providing the ability to detect and diagnose cancer with unprecedented precision and speed.

For example, Paige’s PanCancer Suite, developed using Paige Virchow V2, one of Paige’s state-of-the-art foundation models trained on 3 million digitized slides and 1.8 billion parameters, recently demonstrated clinical grade ability to help pathologists detect cancer in more than 40 tissue and organ types, including the most subtle complexities of common, rare, and precancerous regions in multiple tissue and organ types. It prioritizes cases, flags critical findings, and serves as a quality control tool to ensure no cancer was missed before signing off.

Another example, Paige OmniScreen is extremely promising for its ability to simultaneously screen 1600 molecular biomarkers without needing to re-biopsy patients, providing insights from routine H&E-stained digital pathology slides into the genomic makeup and profile of cancer. OmniScreen recognizes key mutations and pathways critical to cancer progression.

This helps pharmaceutical companies develop new drugs targeting genomic signals that may not have been identified before or as easily identified. And, once identified, these biomarkers can be used to screen patients for enrollment into clinical trials to bring drugs to market faster and prevent undue harm. This breakthrough represents a major leap forward in the accuracy, speed, and cost-effectiveness of cancer diagnosis and treatment selection.

What Ethical Considerations Should Be Addressed when Implementing AI Solutions in Healthcare, Particularly in Areas such as Patient Data Privacy and Decision-Making?

It is important to consider how AI systems are trained and assess potential biases. One of the best ways to combat bias is by using large, high-quality, and representative datasets that account for variability. Only with sufficient data diversity can AI models accurately reflect real-world scenarios and avoid biases—such as incorrectly identifying cancer due to a lack of exposure to certain tissue types.

Building trust in AI is not easy. Clinical validation, peer-reviewed independent studies, and real-world usage build credibility, alongside ongoing product support to maintain consistent performance.

The capabilities of Paige’s Virchow Foundation Models were highlighted in a recent publication in Nature Medicine. And, for about five years before any competitor, Paige was the only company in this space with FDA approval—largely due to the extensive, representative, and unbiased data used in its models. Paige’s latest foundation models were trained on de-identified data from more than 225,000 patients, representing a broad spectrum of gender, race, ethnicity, and geographic diversity. This diversity ensures that the technology is generalizable and equitable, reducing disparities in diagnosis and meeting the highest regulatory standards for ethical AI in healthcare.

It is also important to note that these models support pathologists, who maintain ultimate decision-making authority, ensuring that AI enhances rather than replaces human expertise. By combining AI-driven insights with pathologists’ knowledge, the technology democratizes access to precision healthcare, reinforces clinical decisions, and uncovers patterns invisible to the human eye—ultimately improving patient outcomes.

Ensuring AI is used ethically in healthcare requires transparency, oversight, and a commitment to responsible implementation. AI should support—not dictate—medical decisions, preserving physician autonomy while elevating diagnostic accuracy. By prioritizing representative data, rigorous validation, and human oversight, AI can be developed that is both effective and ethically sound earning the trust of patients, providers, and regulatory bodies.

What Challenges do Healthcare Organizations Face when Integrating AI into their Existing IT Infrastructure, and How Can These Be Mitigated?

In the example of implementing AI in pathology, there are significant, valuable opportunities to enhance patient care and streamline workflows, but also critical challenges to address. Labs require adequate space, hardware to digitize slides, and robust computational infrastructure to support AI operations, which requires capital investment. However, the biggest hurdle in AI adoption is change management.

While many laboratories recognize the potential of digital pathology, adoption remains gradual due to concerns around digitization costs, logistical barriers, and the complexity of integrating AI into established workflows. Overcoming these hurdles requires both investment and a shift in mindset, as pathologists transition from traditional glass slides and microscopes to AI-driven tools.

AI and digital pathology align with the quadruple aim of healthcare—reducing costs, improving clinical outcomes, and enhancing both physician and patient experience. As adoption grows, the efficiency and accuracy AI provides will continue to improve outcomes for both patients and pathologists.

To drive adoption, Paige prioritizes education and collaboration, equipping pathologists and laboratory professionals with the knowledge of how to effectively integrate AI into their work.

Looking ahead, a new generation of informatics-driven pathologists is emerging, one that is highly skilled in leveraging AI and curating its insights. This shift parallels other functional areas in healthcare where AI adoption is accelerating. As pathologists engage with whole-slide imaging and the depth of insights available digitally, they will find their work more manageable, efficient, and impactful, ultimately leading to better patient care.

How Can AI-Driven Healthcare Solutions be Made Accessible and Equitable Across Diverse Populations, Considering Disparities in Technology Adoption?

Ensuring accessibility and equity in AI-driven healthcare starts with high-quality, representative data. By training AI tools on diverse datasets, these solutions can more accurately serve a wide range of demographic groups.

Paige recently open-sourced several of its foundation models, including Virchow and PRISM, to foster broader innovation in cancer care. By making these high-quality models openly available, Paige empowers healthcare providers, AI developers, and researchers to build upon cutting-edge technology. This initiative expands access to more accurate and equitable cancer diagnostics, helping drive meaningful advancements in patient care worldwide.