OASIS: A New AI-Powered Tool for Measuring Meibomian Gland Health

Imagine cutting meibomian gland analysis from 20 minutes to just 3. That’s the power of OASIS.

If you’ve ever treated or studied dry eye disease (DED), you know that meibomian gland dysfunction (MGD) is the main culprit, responsible for over 85% of cases.¹ These tiny glands in the eyelid are essential for keeping the eye’s surface lubricated, and when they atrophy, patients end up with evaporative dry eye.

The problem? Measuring gland health is tricky. Current methods rely on clinicians estimating gland loss by eye, which is time-consuming, subjective and often inconsistent.

Meet OASIS

That’s the gap the Ophthalmic Segmentation and Analysis Software (OASIS) was designed to fill. Published in Translational Vision Science & Technology (TVST), OASIS is an interactive image editor that blends manual annotation with deep-learning–assisted segmentation.² It offers:

Deep-learning–assisted segmentation. A trained model automatically infers gland boundaries, which can then be refined.
Manual flexibility. Tools like pens, erasers and super-pixel overlays make annotation fast and precise.
Built-in enhancement filters. Contrast-limited adaptive histogram equalization (CLAHE) improves gland visibility instantly.
Three-mask approach. Clinicians can annotate eyelid, glands and gland-loss regions—something most other tools don’t support together.
Instant metrics. OASIS calculates gland loss area, gland count, percent gland loss, gland area and even Pult meiboscale grades on the spot.

The goal is to enable objective, reproducible and efficient analysis of meibography images for research and clinical practice.

Inside the OASIS Study

A team of researchers conducted a natural history study with 2,439 meibography images from 325 patients across 11 clinical sites in the United States. Each patient was imaged at baseline and again 90 days later using Johnson & Johnson Vision (California, USA) LipiView II devices. Both upper and lower eyelid images were analyzed.

OASIS allowed clinicians to annotate images manually by marking the eyelid, glands and gland loss areas, or to use deep-learning assistance to generate gland masks that could then be refined.

The software also included features like preset filters for enhancing gland contrast and super-pixel overlays to speed up annotation. By combining automation with flexible editing, OASIS offered both efficiency and accuracy.

Results that stand out

When compared to traditional manual methods, OASIS delivered striking improvements.

Speed. Reduced analysis time from 15 to 20 minutes to under three minutes, an 87% time savings.
Accuracy. Automated segmentation achieved a Dice score of 0.84 ± 0.07 against clinician annotations.
Consistency. OASIS and clinician-assigned Pult grades showed substantial agreement (Cohen’s kappa = 0.79).
Scalability. More than 1,500 images were processed automatically with robust performance, even on real-world, imperfect images.

Why this matters for MGD

The ability to quickly and objectively measure gland loss and morphology could be a game-changer for MGD diagnosis and management. Instead of relying on subjective grading, clinicians could use reproducible, quantitative data to track subtle progression across visits, standardize grading across practices and studies and free up time for patient care instead of manual tracing.

The bigger picture

The tool isn’t perfect yet. Image quality and certain gland morphologies still pose challenges. But OASIS points toward a future where AI-supported, interactive platforms become the norm for ocular imaging. More accurate tracking of gland changes could transform how we diagnose, monitor and treat dry eye disease.

But OASIS isn’t just about counting glands. The study highlights how the software could be expanded to measure additional morphologies such as tortuosity, thickening or ghost glands. Future versions may integrate even more sophisticated AI segmentation to reduce manual corrections.

For now clinicians have a glimpse of a future where gland analysis is quicker, sharper and (dare we say) almost painless.

Editor’s Note: This content is intended exclusively for healthcare professionals. It is not intended for the general public. Products or therapies discussed may not be registered or approved in all jurisdictions, including Singapore.

References

Sheppard JD, Nichols KK. Dry eye disease associated with meibomian gland dysfunction: focus on tear film characteristics and the therapeutic landscape. Ophthalmol Ther. 2023;12(3):1397–1418.
Joseph N, Shivade V, Chen J, et al. Ophthalmic segmentation and analysis software (OASIS): A comprehensive tool for quantitative evaluation of meibography images. Translational Vision Science & Technology. 2025;14:22.