Sign in Subscribe
Machine learning

Machine Learning Enhances Early Detection of Keratoconus in Multicenter Study

Strategic Insights

— 1 min read

Researchers at Peking University Third Hospital have developed a machine learning diagnostic system that integrates Scheimpflug tomography, corneal biomechanics, and clinical risk factors to improve early detection of keratoconus. The prospective, multicenter, cross-sectional study involved 1,035 eyes from 1,035 participants across five centers between 2020 and 2024. The machine learning model achieved high accuracy in detecting forme fruste keratoconus, subclinical keratoconus, and clinical keratoconus, outperforming other models and indices.

Key Takeaways:

  • The machine learning model integrated Scheimpflug tomography, corneal biomechanics, and clinical risk factors to enhance early detection of keratoconus.
  • The study involved 1,035 eyes from 1,035 participants across five centers between 2020 and 2024.
  • The machine learning model achieved high accuracy in detecting forme fruste keratoconus, with an area under the receiver operating characteristic curve (AUROC) of 0.975.
  • The model outperformed combined-feature (AUROC = 0.963), CRF-only (AUROC = 0.856), and device-only models (AUROC = 0.885) in the test set.
  • The multimodal machine learning system improved early keratoconus detection, particularly for forme fruste keratoconus.
  • The study concluded that this approach may enhance clinical decision-making and screening for refractive surgery candidates.

Statistics:

  • 1,035 eyes from 1,035 participants were involved in the study.
  • The machine learning model achieved an AUROC of 0.975 for forme fruste keratoconus detection.
  • The model outperformed combined-feature, CRF-only, and device-only models in the test set.
  • The machine learning model achieved AUROCs of 0.991 and 1.000 for subclinical and clinical keratoconus detection, respectively.
  • The study involved five centers across a four-year period.

Sources:

  • American Journal of Ophthalmology, Enhancing Early Keratoconus Detection with Multimodal Machine Learning: Integrating Tomography, Biomechanics, and Clinical Risk Factors, American Journal of Ophthalmology, 2025, Elsevier Science Inc, Ste 800, 230 Park Ave, New York, NY 10169, USA.
  • NewsRx, New Keratoconus Study Results from Peking University Third Hospital Described, Medical Devices & Surgical Technology Week, September 21, 2025, p 1613.

Read more