Myopia treatment affects meibomian gland morphology in pediatric patients treated with orthokeratology or soft contact lenses, according to research published in Contact Lens & Anterior Eye. Artificial intelligence (AI) may be instrumental in helping clinicians detect these changes, the report shows.
Researchers retrospectively reviewed data from children with myopia who were treated with either orthokeratology (ortho-k; n=89; 36% girls; mean age, 11.40 years) or soft contact lenses (SCL; n=70; 44.3% girls, mean age, 10.94 years). The team used an AI analytic system to assess how the 2 myopia treatments affect meibomian gland tortuosity, height, width, density, and vagueness values and obtained tear meniscus height (TMH), noninvasive tear breakup time (NIBUT), and meibography using a corneal topography system.
At follow-up (mean, 20.8 months), both myopia control treatments resulted in significant increases in meibomian gland width of the upper eyelid (P =.031 and P =.011 for ortho-k and SCL, respectively) and significant decreases in meibomian gland vagueness values for both the upper and lower eyelids (P ≤.023 for all). Additionally, ortho-k myopia treatment affected meibomian gland tortuosity of the upper eyelid, resulting in a significant increase at follow-up (P <.001). No significant changes in TMH or NIBUT were noted with either treatment (all P > .05).
Overall, the AI model showed that ortho-k treatment positively affected meibomian gland tortuosity of both upper and lower eyelids and width of the upper eyelid, while negatively affecting meibomian gland density of the upper eyelid and vagueness values of both upper and lower eyelids, according to the report.
The report shows that SCL myopia treatment affects meibomian gland width of both upper and lower eyelids, the meibomian gland height of the lower eyelid, and tortuosity of the upper eyelid positively, while negatively affecting meibomian gland vagueness value of both upper and lower eyelids.
The investigators found no association between treatment duration and TMH, NIBUT, and meibomian gland morphological parameters among participants treated with ortho-k, but SCL treatment duration negatively affected meibomian gland height of the lower eyelid (P = .002).
“AI systems based on deep learning can significantly improve the accuracy of meibography images analysis …,” according to the researchers. “Simultaneously, an AI system could solve the difficulty of precisely determining when the MG dropout grade is near the grading transition limit while eliminating the subjective bias of observers. Then, an analysis may occur automatically and be accurately expressed with novel quantitative indicators, such as MG height, width, tortuosity, density, and vagueness value.”
Study limitations include a failure to include a control group composed of individuals who were not treated with contact lenses, failure to consider confounding factors such as electronic screen use and lens materials, and failure to examine the clinical relevance of these findings.