Medical startup Paige unveils Alba AI all-in-one assistant for pathology research

Even those of outside the field of medicine have some awareness of how complex, sprawling, and data-driven the various disciplines within it are.

Take pathology — the science of diagnosing disease and injury by analyzing body tissue. Simply looking at human tissue and trying to understand what’s going wrong with it is a tremendous challenge in its own right, but pathologists must also access patient’s medical records to better understand their unique background and lifestyle in order to make a fully informed, accurate diagnosis. And further, they must then translate their findings into readable reports for other doctors — such as clinicians (those who treat patients directly). And, they frequently must follow-up on their findings with actions, such as ordering stains or preparing for tumor boards.

Doing all this is of course what most pathologists agree to sign up for when undertaking their studies and jobs, but with modern technology, specifically generative AI models, they have the opportunity to be more efficient and spend more time using their mind rather than translating information mechanically across domains.

Enter Paige, a 7-year-old medical technology startup headquartered in New York City that aims to transform cancer research, diagnosis and treatment with its proprietary, in-house AI tools and models designed for pathologists — though it is starting with a tool for internal research usage at medical facilities only, not treatment.

That tool is called Alma, an AI assistant or copilot that allows pathologists to simply type in questions in natural language queries on their work computers and immediately have a wealth of information available to them about a particular patient, extracted from their official medical records, as well as help them prepare reports and take their follow-up actions.

As Dr. Juan Retamero, VP of Clinical Diagnostics at Paige, told VentureBeat in a video call today: “Most AI solutions focus on just one aspect—image analysis—but our models, including Alba, are designed to help with all three.”

What Paige’s new Alba AI copilot offers

Paige Alba consolidates patient data from multiple disparate sources such as Electronic Health Records (EHRs), Laboratory Information Systems (LIS), and Image Management Systems (IMS).

These hospital-based data repositories contain critical patient details, pathology reports, radiology findings, and historical medical information. By aggregating this diverse data into a single system, Alba eliminates the need for medical professionals to manually navigate through various platforms, reducing administrative overhead and allowing pathologists to focus on more critical tasks.

The AI-powered system summarizes and presents patient history, prior pathology reports, radiology findings, and other key data, offering pathologists actionable insights within seconds.

Alba further enhances decision-making by leveraging Paige’s portfolio of clinical-grade AI tools, including Paige Omniscreen, which screens molecular biomarkers in tissue samples to help identify suspicious areas for potential cancer. The AI system generates interim case evaluations for expert review, streamlining the process of generating diagnostic reports. Physicians can review, modify, and approve these reports, all via voice command, enhancing both the speed and efficiency of the diagnostic workflow.

Retamero emphasized that “Alba combines visual analysis with natural language processing, so instead of just identifying cancer on a slide, it also helps pathologists by writing structured reports and pulling relevant clinical data from electronic health records or radiology systems.”

This holistic approach enables Alba to not only assist in diagnosis but also manage the administrative workload, reducing the time pathologists spend on repetitive tasks.

Proprietary in-house AI foundation models trained on millions of medical images — largest for pathology

While Alba represents the latest development, it builds on Paige’s extensive work in AI-powered cancer diagnostics. In August 2024, the company announced its second-generation Virchow models—Virchow2 and Virchow2G. These models are part of Paige’s million-slide foundation model and have been developed in collaboration with Microsoft, using one of the largest and most diverse datasets in clinical pathology.

The Virchow2 and Virchow2G models were trained on over 3 million pathology slides collected from more than 800 labs across 45 countries, representing a wide range of patient demographics, including gender, race, ethnicity, and geographic locations.

The dataset encompasses over 225,000 patients and includes over 40 different tissue types stained with H&E and diverse immune-stains (IHC). This comprehensive and diverse dataset allows the AI to deliver deeper insights into cancer across a broad spectrum of pathology use cases, enabling the models to assist in rapid and accurate diagnosis, even in complex or rare cases.

The sheer scale of this data, alongside the models’ 1.8 billion parameters, makes Virchow2G the largest AI model ever created for pathology.

Retamero explained that Paige’s access to this extensive dataset through “collaboration with Memorial Sloan Kettering Cancer Center” in New York City gives the company a significant edge in developing highly effective AI tools.

This comprehensive archive of pathology data allows Paige to train models that are not just advanced in theory, but capable of delivering meaningful, real-world clinical insights.

The introduction of Paige Alba builds on this robust foundation by integrating the insights generated by these advanced models into real-time clinical use. Together, Alba and Virchow2 represent Paige’s comprehensive approach to cancer care—ranging from diagnostics to research, all driven by AI.

Research-only for now

It’s important to note that while Alba promises to revolutionize clinical workflows, it is currently designated for research use only (RUO) and is not yet approved for diagnostic procedures.

However, its potential for improving diagnostic accuracy, particularly in oncology, signals a strong future for AI applications in clinical environments, and could pave the path for similar apps designed to treat patients. For now, it will be used to help pathologists research overall cancer features and better understand the disease.

Paige’s ultimate goal is to push the boundaries of what AI can achieve in healthcare. The company’s foundation models, like Virchow2 and Virchow2G, have already demonstrated the impact that large-scale AI can have on cancer diagnostics. By continuing to innovate, Paige is moving closer to a future where AI not only aids in cancer detection but also enhances personalized treatment plans.

According to Yousfi, Alba’s introduction is just the beginning. As AI technologies evolve, Paige plans to further integrate its capabilities into clinical practice, ensuring that medical professionals have access to the best tools for diagnosing and treating cancer.