One way cancer specialists detect the disease is by examining cells and bodily fluids under a microscope, a time-consuming and labor-intensive process called cytology. It involves visually inspecting tens of thousands to one million cells per slide for subtle 3D morphological changes that might signal the onset of cancer. But AI offers an approach that is potentially faster and more accurate.
In a new study published in the journal Nature, researchers demonstrate an AI-powered 3D scanning system that can automatically sort through samples and identify abnormal cells with performance approaching that of human experts.
Building digital models
The team developed a system called Whole-Slide Edge Tomography, which uses a scanner to capture a series of images at different depths to create a 3D digital model of every cell on a slide.
The system’s AI program then identifies individual cells and analyzes their 3D shape and internal details to determine whether they are healthy or abnormal. It then organizes this information using an approach the researchers named Cluster of Morphological Differentiation (CMD).
This plots the cells on a map showing the locations of both healthy cells and those shifting toward disease. It allows doctors to see the big picture of a patient’s health at a glance, rather than having to hunt for a few abnormal cells among millions of healthy ones.
The scientists tested their platform on hundreds of cervical samples. The AI was highly accurate in identifying different stages of the disease. For example, it achieved AUCs (a statistical measure of how well a system distinguishes between healthy and diseased cells, where 1 is perfect) of 0.84 for early-stage changes and 0.89 for more advanced phases.
The researchers then expanded the test to a larger group of 1,124 slides collected from four different medical centers. Its AUC values ranged from 0.86 to 0.91 for lower-grade abnormalities and reached 0.97 for high-grade lesions.
The future of cancer diagnostics?
The speed of the analysis was just as impressive, taking only a matter of minutes to process an entire slide. At the individual cell level, the AI achieved near-perfect performance in distinguishing between healthy cells and abnormal ones. The system also found abnormal cells in samples that had originally been labeled as normal by human experts.
“Our platform establishes a scalable, real-time cytology pipeline with clinical-grade autonomy and lays the foundation for an objective, reproducible and discovery-driven diagnostic paradigm,” wrote the team in their paper.
The next step for the team is to expand this platform beyond cervical cancer to test its effectiveness at screening for other types of the disease. https://medicalxpress.com/news/2026-02-faster-cancer-screening-ai-abnormal.html
Recent Comments