When Aung Naing's team at MD Anderson Cancer Center in Houston started using artificial intelligence to study tumor biopsies, they weren't sure what they'd find. What they discovered could change how doctors treat rare cancers. Patients whose AI scans showed favorable patterns lived nearly four times longer on average than those without them—42 months versus just 10 months. That's a difference of more than two and a half years, spent with family instead of a hospital bed. The findings appeared in the Journal for ImmunoTherapy of Cancer.
Immunotherapy harnesses the body's own immune system to fight cancer. But predicting which patients will respond to it has always been difficult, especially with rare cancers that doctors see less often and understand less well. Naing, a professor of Investigational Cancer Therapeutics at MD Anderson, previously identified two clues hidden in tumor biopsies that could signal a good response: how many immune cells were already present inside the tumor before treatment began, and how those immune cells changed as treatment progressed.
The problem? Manually counting thousands of cells on pathology slides takes enormous time and effort. This is where AI steps in. The new tool analyzes standard tumor slides—the same ones hospitals already collect during routine biopsies—and rapidly measures these features across multiple samples from the same patient over time. No special new tests required.
The results were striking. Patients showing both favorable signals—increased immune cell infiltration combined with shrinking tumor content—had a 64 percent lower risk of their cancer progressing or of dying. Naing called the approach an "exciting step forward."
"AI-based pathology has the potential to provide clinicians with useful information on both the tumor and its surrounding microenvironment, helping to guide personalized treatment decisions for patients receiving immunotherapy," he said.
Before this tool reaches hospital waiting rooms, researchers need to test it on larger groups of patients to confirm the findings hold up. But the early signs suggest something important: meaningful health insights may already be hiding in routine medical tests that doctors use every day, waiting for the right technology to reveal them.