Compact AI-Powered Microscope Enables Rapid Cost-Effective Cancer Scoring

However, conventional whole-slide imaging scanners are prohibitively expensive, costing over USD 100,000, and their large size limits use in many labs and resource-limited settings. To address these challenges, researchers have now developed a compact scanning microscope that uses artificial intelligence (AI) to extract diagnostic value from blurred tissue images at a fraction of the cost.

BlurryScope, a portable, cost-effective scanning microscope, has been designed by researchers at UCLA (Los Angeles, CA, USA) for HER2 scoring in breast cancer tissue samples. Built for under USD 650 and weighing just 2.26 kg, the device measures 35 x 35 x 35 cm, making it highly accessible. Unlike conventional microscopes that capture sharp still images, BlurryScope continuously scans slides, producing motion-blurred images that are then analyzed by a deep neural network trained to classify HER2 expression.

In blinded experiments with 284 patient tissue cores, the system achieved nearly 80% accuracy across standard HER2 scoring categories and almost 90% accuracy when grouped into clinically actionable categories. Repeated scans of the same samples demonstrated reproducibility, with more than 86% of classifications matching across multiple runs. The results, published in npj Digital Medicine, validate both the robustness of the AI and the reliability of the device for clinical application.

BlurryScope automates the workflow from scanning to classification, enabling rapid triage, preliminary assessments, and use in clinics where high-end systems are impractical. Beyond HER2 scoring, the same AI-powered approach could extend to other tissue stains, biomarker analyses, and even different imaging modalities. By co-designing optical systems and AI algorithms, the technology demonstrates a pathway toward simplifying biomedical imaging hardware while broadening diagnostic access worldwide.

“With BlurryScope, we are redefining what affordable microscopy can look like,” said Dr. Aydogan Ozcan, Chancellor’s Professor of Electrical and Computer Engineering at UCLA. “By harnessing deep learning to interpret motion-blurred images, we can deliver clinically meaningful HER2 scoring at a fraction of the cost and size of conventional pathology scanners.”

Related Links:
UCLA