SuperμMAPPS: a supervised AI-phasor-based method to investigate collagen micro-architecture and early detect tumorous tissue in human samples

Despite their key-role during the histopathological diagnosis, staining procedures are expensive and time-consuming. Label-free microscopy provides an alternative since it allows the visualization of endogenous proteins without the need of extrinsic dyes. SuperµMAPPS, a novel AI-based method, analyzes the Polarized Second Harmonic Generation signal from collagen to characterize its micro-architecture in terms of fibrils mean orientation θF and anisotropy γ, related to tumor development. After a proper validation on synthetic images, human breast cancer samples at different growth stages have been analyzed through SuperµMAPPS, highlighting its capability to detect tumorous tissue at early stages in a real clinical context.