ESCRS - PP20.09 - Blinking Pattern Analysis Via Deep Learning System As A Diagnostic Tool For Tear Film Instability In Children With Long-Term Use Of Orthokeratology

Blinking Pattern Analysis Via Deep Learning System As A Diagnostic Tool For Tear Film Instability In Children With Long-Term Use Of Orthokeratology

Published 2025 - 43rd Congress of the ESCRS

Reference: PP20.09 | Type: Poster | DOI: 10.82333/rffw-w930

Authors: Siyuan Wu* 1 , Yue Wu 1 , Yinghai Yu 2 , Bilian Ke 1

1Ophthalmology,Renji Hospital, Shanghai Jiao Tong University School of Medicine,Shanghai,China, 2Ophthalmology,Shanghai General Hospital, Shanghai Jiao Tong University School of Medicine,Shanghai,China;School of Health Science and Engineering, University of Shanghai for Science and Technology,Shanghai,China

Purpose

To investigate the diagnostic value of deep learning-based blinking pattern analysis for detecting tear film instability in children with long-term orthokeratology (ortho-K) use, focusing on its potential application in early identification of keratoconjunctivitis sicca and contact lens-associated dry eye disease.

Setting

All patients involved were assessed and managed locally by clinical members of the refraction team at Shanghai General Hospital affiliated to Shanghai Jiao Tong University School of Medicine.

Methods

 

  • Design: Retrospective case-control study (31 ortho-K users [58 eyes] vs 31 controls[58 eyes]).
  • Intervention:
    • Assessments:
      • Ocular surface evaluation: Keratograph 5M, LipiView, OSDI scoring
      • Tear film stability: First/average non-invasive breakup time (NIBUT)
    • Quantitative blink analysis: U-Net-Swin-Transformer model processing 20s blink videos to observe:
      • The frequency of incomplete blinks (IB) and complete blinks (CB)
      • Incomplete blink ratio (IBR)
      • Interpalpebral height% (IPH%, defined as IPH/image height ratio)
      • Relative IPH% (defined as mean/max IPH% ratio), proposed to indicate the extent of incomplete blink
      • The duration of the closing, closed, and opening phases in the blink wave
    • Statistics: Independent T-test; Mann-Whitney test; Spearman correlation.

Results

 

  • Model performance: 98.1% blink classification accuracy.
  • Ortho-K abnormalities (vs controls):
    • Reduced NIBUT: First-NIBUT (11.75 ± 7.42s vs 14.87 ± 7.93s, p=0.030); Average-NIBUT (13.67 ± 7.0s vs 16.60 ± 7.24s, p=0.029)
    • Elevated incomplete blinking: IBR (0.81 ± 0.28 vs 0.46 ± 0.39, p<0.001); Relative IPH% (0.3229 ± 0.1539 vs 0.2233 ± 0.1960, p=0.004)
    • Prolonged eye-closing phase (0.18 ± 0.08 s vs 0.15 ± 0.07 s, p = 0.032) and opening phase (0.35 ± 0.12 s vs 0.28 ± 0.14 s, p = 0.015)
  • Diagnostic correlation: Significant negative relationship between the frequency of incomplete blinks and NIBUT (first-NIBUT: r = −0.292, p = 0.004; avg-NIBUT: r = −0.351, p < 0.001).

Conclusions

Deep learning system-based blink pattern analysis objectively identifies tear film instability in long-term ortho-K users, providing a novel diagnostic marker for subclinical keratoconjunctivitis sicca. Incomplete blinks, prolonged eye-closing and opening phase were the main characteristics of their blink patterns compared to the healthy control. More incomplete blinks were associated with reduced NIBUT, suggesting that less effective blinking may contribute to tear film instability. Monitoring IBR and IPH% could enhance early detection of contact lens-associated dry eye disease in children undergoing ortho-k treatment in myopia control. Integration of this tool into routine follow-up may optimize contact lens safety in myopia control.