If you teach AI to look at medical scans like they're ripples of light, it gets way better at spotting cancer—no matter what hospital gear you’re using.
arXiv · March 18, 2026 · 2603.15980
The Takeaway
AI models often fail when moved to a new hospital because different scanners produce slightly different image qualities. By applying the laws of optical physics to 'standardize' these images as if they were light waves passing through a lens, researchers boosted cancer detection accuracy from 70% to over 90%.
From the abstract
Deep learning has achieved remarkable success in medical image analysis, yet its performance remains highly sensitive to the heterogeneity of clinical data. Differences in imaging hardware, staining protocols, and acquisition conditions produce substantial domain shifts that degrade model generalization across institutions. Here we present a physics-based data preprocessing framework based on the PhyCV (Physics-Inspired Computer Vision) family of algorithms, which standardizes medical images thr