Machine learning shows promise in estimating tear osmolarity for contact lens wearers

There is potential for integrating machine learning techniques into contact lens research and practice to enhance the measurement and understanding of tear osmolarity, a crucial parameter associated with contact lens-induced dry eye, according to a study.

The study demonstrates that traditional methods may be insufficient for accurately predicting tear osmolarity, but advanced machine learning models show promise in achieving higher accuracy levels.

The research utilized data from 175 participants, primarily healthy subjects eligible for soft contact lens wear. Various clinical assessments were performed, encompassing symptom evaluation, tear film parameters, ocular surface health, and Meibomian gland assessment.

The study highlighted the limitations of traditional linear regression techniques in predicting tear osmolarity accurately. However, employing more sophisticated regression models yielded promising results, explaining approximately 32% of the variability in tear osmolarity. Key predictors identified included parameters such as non-invasive keratometric tear film break-up time (NIKBUT), tear meniscus height (TMH), ocular redness, Meibomian gland coverage, and symptom assessment questionnaires.

In classification tasks aimed at categorizing tear osmolarity levels, the machine learning models achieved an accuracy of approximately 80%. This classification, differentiating between low, medium, and high osmolarity levels, underscored the significance of parameters such as NIKBUT, TMH, ocular redness, Meibomian gland coverage, and symptom assessment.

Reference
Garaszczuk IK, Romanos-Ibanez M, Consejo A. Machine learning-based prediction of tear osmolarity for contact lens practice. Ophthalmic Physiol Opt. 2024;doi: 10.1111/opo.13302. Epub ahead of print. PMID: 38525850.