Related Experiment Video
Updated: Aug 4, 2025
07:51
Full-Field Optical Coherence Microscopy for Histology-Like Analysis of Stromal Features in Corneal Grafts
Published on: October 21, 2022
1.7K
Comparison of machine learning-based algorithms using corneal asymmetry vs. single-metric parameters for keratoconus
Gaurav Prakash1, Chandrashan Perera2, Vishal Jhanji3
1Department of Ophthalmology, University of Pittsburgh School of Medicine, Pittsburgh, PA, USA.
Summary
Corneal asymmetry ratios effectively diagnose keratoconus, outperforming traditional metrics like maximum anterior curvature (Kmax) and thinnest corneal thickness. Machine learning models utilizing these ratios achieved high sensitivity and specificity for keratoconus detection.
Area of Science:
- Ophthalmology
- Medical Imaging
- Machine Learning
Background:
- Keratoconus is a progressive corneal ectasia leading to vision impairment.
- Accurate diagnosis relies on identifying subtle changes in corneal shape and thickness.
- Conventional parameters like Kmax and thinnest corneal thickness have limitations in early detection.
Purpose of the Study:
- To compare the diagnostic performance of corneal asymmetry parameters against standard keratoconus diagnostic metrics.
- To evaluate the efficacy of machine learning models in identifying keratoconus using these parameters.
Main Methods:
- Retrospective case-control study involving 290 keratoconus eyes and 847 normal eyes.
- Corneal tomography data acquired using Scheimpflug technology.
- Machine learning models developed using Python libraries (sklearn, FastAI) trained on topography metrics and clinical diagnoses.
Main Results:
- Corneal asymmetry ratios demonstrated superior diagnostic performance compared to Kmax and thinnest pachymetry.
- Machine learning models using only corneal asymmetry ratios achieved 99.0% sensitivity and 94.0% specificity.
- This approach outperformed models using combined traditional measures (Kmax, thinnest cornea, inferior-superior asymmetry).
Conclusions:
- Corneal asymmetry ratios, when analyzed by machine learning, provide a highly sensitive and specific method for keratoconus diagnosis.
- This parameter set shows promise for improving early detection and management of keratoconus.
- Further validation with larger datasets and diverse populations is recommended.
