Development of a microscopy-based diagnostic test for alkali injury-induced limbal stem cell deficiency through...

Diagnosing limbal stem cell deficiency (LSCD) remains challenging, and this study evaluates a newly developed Autofluorescence Multispectral Imaging (AFMI) technology for its diagnostic potential. Unstained fixed corneal cross-sections from 36 mice were used in this study. The right eyes of these mice had alkali-induced LSCD, while the left eyes served as healthy controls. Autofluorescence multispectral images of the corneal epithelium were acquired across multiple spectral channels to track changes in endogenous fluorophore spectra. Big data analytics and machine learning classifiers were applied to differentiate LSCD-affected from healthy tissues. The performance of this technique was evaluated by histological evaluation. Imaging was conducted on both the Operetta CLS (PerkinElmer) for research applications and a custom-modified Olympus IX83 system designed for potential clinical translation. Significant differences in autofluorescence signal intensities (p < 0.0001) were observed between diseased and control tissues across several key fluorophores. The machine learning classifier achieved high prediction accuracy using both the Operetta CLS system (90.38 %) and a custom-modified Olympus IX83 platform (83.28 %) during 10-fold cross-validation, and the false-color prediction maps showed strong agreement with histological assessments. These results highlight new perspectives on LSCD through the characterization of autofluorescent biomarkers and demonstrate the feasibility of AFMI as a diagnostic tool in a mouse model, providing a basis for future investigations into its possible adaptation for non-invasive clinical assessment of LSCD and other ocular surface disorders.