Artificial Intelligence (AI) could help an op-tometrist analyze the tear film and meibomian glands to assist in diagnosis. This, in turn, could guide the treatment of dry eye disease (DED).
Here, I discuss the types of AI involved and the specific technologies in development related to tear film analysis and meibomian gland imaging and monitoring.
AI Benefits
In the case of DED, AI algorithms and tools are in development that classify large sets of clinical images into different categories based on the characteristics of diseases. As a result, these technologies, through their supercomputing power and data-mining ability, can assist optometrists in clinical decision-making1 and have the propensity to help standardize assessments across different practitioners. This would ensure consistency in diagnosis regardless of individual training backgrounds.2

Tear Film
Imaging systems are in development that have built-in algorithms based on more than one million clinical images. These systems could provide ODs with detailed feedback on ocular surface markers to aid them in detecting tear film composition, tear meniscus height, corneal morphology, and blinking abnormalities.1
With regard to tear film in-stability, different AI systems that have deep learning (DL) algorithms are correctly identifying reduced tear breakup time and assessing the risk factors in predicting tear film instability.3
When it comes to corneal staining, a new AI algorithm can analyze the staining patterns from photos of patients who have Sjögren’s syndrome (SS) and ocular graft-versus-host disease (GVHD). Additionally, the AI analysis of the images has shown good correlation with experts’ scores of staining patterns. What’s more, the algorithm can distinguish between SS and GVHD patients based on staining patterns with good specificity and sensitivity.4
Meibomian Glands
Diagnostic devices, from slit lamp cameras, to stand-alone meibographers, are using DL AI algorithms to provide gland grading. One such technology assessed 209 meibomian gland images and achieved a 95.6% grading accuracy on average.1 Models for both obstructive and atrophic meibomian gland dysfunction had a greater than 88% sensitivity and a greater than 95% specificity.1
Potential Efficiency
Based on the available research, AI has great potential in improving the efficiency of diagnosing ocular surface diseases, as it can assist the optometrist with objective clinical decisions by comparing new images to a library of normative data.1 It may also have the capacity to build a foundation for the accurate treatment of patients.5 That said, it’s important to note that the use of AI in this area is still in its early stages, so further research is needed to fully elucidate its impact. Such research includes addressing ethical concerns, such as data privacy and bias in AI algorithms, to ensure that the technology is used in a responsible and equitable manner.5 I think that being open to new technologies, while understanding their limitations is essential to treating our patients effectively.
References
1. Zhang Z, Wang Y, Zhang H, et al. Artificial intelligence-assisted diagnosis of ocular surface diseases. Front Cell Dev Biol. 2023;11:1133680. Published 2023 Feb 17. doi:10.3389/fcell.2023.1133680
2. Swatts S. The Top Ten Reasons to Use AI in your Dry Eye Practice. Rev Opto Bus. https://reviewob.com/the-top-10-reasons-to-use-ai-in-your-dry-eye-practice/ 2024 Aug 14. Accessed March 5, 2025.
3. Fineide F, Storås AM, Chen X, Magnø MS, Yazidi A, Riegler MA, et al. Predicting an unstable tear film through artificial intelligence. Sci Rep. 2022;12:21416. doi: 10.1038/s41598-022-25821-y.
4. Pellegrini M, Bernabei F, Moscardelli F, Vagge A, Scotto R, Bovone C, et al. Assessment of corneal fluorescein staining in different dry eye subtypes using digital image analysis.
Transl Vis Sci Technol. 2019;8:34. doi: 10.1167/tvst.8.6.34.
5. Pagano L, Posarelli M, Giannaccare G, et al. Artificial intelligence in cornea and ocular surface diseases. Saudi J Ophthalmol. 2023;37(3):179-184. Published 2023 Sep 16. doi:10.4103/sjopt.sjopt_52_23