A Contactless and non-invasive Analysis of Meibomian Gland Cyst Eye Disease Using Hyperspectral Imaging Camera

This work extends our previously published study on meibomian gland cyst (MGC) by introducing a fully contactless, closed-eye hyperspectral imaging (HSI) workflow for detection and severity grading. Images were acquired from 16 subjects and processed via a compact pipeline: region-of-interest extraction, light normalization and denoising, spectral unmixing with MCR-ALS to isolate cyst-related components, and computation of a meibomian-gland-condition level (MGCL) map used for binary detection and five-level grading. On this cohort, the proposed system achieved 93.75% accuracy for binary detection (normal vs. MGC) and 81.25% accuracy for five-class grading, supported by confusion-matrix analysis. Results showed meaningful agreement with independent clinical assessment and standard dryness scoring, while avoiding eyelid eversion and other invasive steps. The pipeline is fast, portable, and suitable for point-of-care screening. Future work will expand the dataset and refine thresholds for broader clinical generalization.