Mount Sinai researchers unveil MARQO, AI tool to accelerate cancer tissue analysis

A team at the Icahn School of Medicine at Mount Sinai has built a computational system that could reshape how pathologists examine cancer tissue. The tool, called MARQO, automates the work of analyzing stained tissue slides—a task that has long demanded hours of manual labor and microscope time. The research, published in Nature Biomedical Engineering, represents a shift toward letting machines handle the heavy sorting work while keeping human judgment at the center of diagnosis.

When a patient is diagnosed with cancer, pathologists stain tissue samples with special markers that highlight immune cells and other biomarkers, then examine them under a microscope to map what's present and how cells are arranged. This process is painstaking. A pathologist might spend hours on a single slide, and even then, the analysis often covers only small sections of the tissue. The work is subjective, time-consuming, and difficult to compare across different labs or studies. MARQO was designed to solve this bottleneck.

Sacha Gnjatic, a professor of immunology and immunotherapy who led the project, explains that the tool fills a gap between what's technically possible and what's practical. Other image analysis systems exist, but they typically require researchers to break slides into smaller patches or rely on expensive computing clusters. MARQO processes entire slides intact and completes the analysis in minutes on standard graphics processing units—the kind of hardware most labs already have. The system works across multiple common staining technologies, which means results from different studies become easier to compare and validate.

The workflow keeps pathologists in control. MARQO automatically identifies cells likely to be positive for specific markers, assigns them coordinates, and measures marker intensity. But it doesn't make the final call. A pathologist reviews the system's work, validates the findings, and interprets what they mean. This hybrid approach—automation plus expertise—is what distinguishes MARQO from fully automated systems that might miss nuance or introduce errors that go unnoticed.

The implications extend beyond research. MARQO is currently designed for laboratory use and has not been validated for clinical diagnostics, but it was built using staining methods that are already standard in pathology clinics. That compatibility means the tool could eventually move into hospital labs, where it might accelerate the discovery of biomarkers that predict which patients will respond to specific treatments. Faster, more consistent analysis could support the development of personalized cancer therapies tailored to individual tumor profiles.

The research team, which includes Mark Buckup, Edgar Gonzalez-Kozlova, Igor Figueiredo, Pauline Hamon, and Giorgio Ioannou, received funding from the National Cancer Institute through multiple grant mechanisms, along with seed support from Mount Sinai itself. The group plans to continue refining the tool—improving its interface, adding spatial analysis features, and scaling it to handle millions of digitized slides in high-performance computing environments. The next phase will test whether MARQO can deliver on its promise to make cancer diagnosis faster, more precise, and more reproducible across institutions.