A new digital biomarker of blepharitis: energy curve of the meibomian edge.

PURPOSE

To develop and validate a novel digital biomarker, the energy curve of the meibomian gland (MG) edge, to assess MG uneven atrophy and aid in diagnosing blepharitis.

METHODS

A retrospective study enrolled 76 dry eye patients (42 with blepharitis, 34 controls). Segmentation of upper eyelid meibography images was accomplished via a convolutional neural network (CNN)-based artificial intelligence (AI) model. The lower margin curve of MGs was extracted using an active contour model (Snake) to compute a composite energy value that integrates elastic, curvature, and smoothness energies. Clinical parameters, including non-invasive tear breakup time (NIBUT), lid margin score, and Meiboscore, were evaluated.

RESULTS

The group showed shorter NIBUT (median: 2.84 vs. 5.18 s, < 0.001) and higher lid margin scores (median: 2 vs. 1, = 0.002) and Meiboscores (median: 1 vs. 1, = 0.009). The group also exhibited significantly higher energy curve values than controls (median: 32.44 vs. 11.20, < 0.001), reflecting pronounced uneven gland atrophy. Meanwhile, MG density significantly influenced energy curve values ( = 0.010). After adjusting for MG density, the energy curve demonstrated strong diagnostic accuracy (AUC = 0.897, sensitivity 78.6%, specificity 91.2%).

CONCLUSION

The energy curve quantifies structural irregularities in MGs caused by infestation, serving as a non-invasive biomarker for early diagnosis. Its integration with meibography enhances clinical workflows, particularly in resource-limited settings.