Related Experiment Video
Updated: Jun 25, 2026
07:51
Full-Field Optical Coherence Microscopy for Histology-Like Analysis of Stromal Features in Corneal Grafts
Published on: October 21, 2022
[Detection of keratoconus based on a neural network with Orbscan]
Murilo Barreto Souza1, Fabrício Witzel de Medeiros, Danilo Barreto Souza
1IMédico Docente da Disciplina de Informática Médica do Curso de Medicina da Faculdade de Tecnologia e Ciências - Salvador (BA) - Brasil. murilobsouza@gmail.com
Arquivos Brasileiros De Oftalmologia
|March 11, 2009
Summary
An artificial neural network accurately identifies keratoconus in Orbscan II tests, aiding clinical diagnosis. This AI tool shows high sensitivity and specificity for detecting corneal abnormalities.
Area of Science:
- Ophthalmology
- Biomedical Engineering
- Artificial Intelligence
Background:
- Keratoconus is a progressive corneal ectasia.
- Accurate diagnosis of keratoconus is crucial for timely intervention.
- Orbscan II topography provides data for corneal analysis.
Purpose of the Study:
- To evaluate an artificial neural network (ANN) for identifying keratoconus.
- To assess the ANN's accuracy in classifying normal versus keratoconus corneas using Orbscan II data.
Main Methods:
- Retrospective analysis of 98 Orbscan II tests from 59 subjects.
- Development and training of an ANN using Java Neural Network 1.1 software.
- The ANN was trained on 73 tests and validated on 25 tests (19 normal, 6 keratoconus).
Main Results:
- The backpropagation method trained the ANN to a 5% error rate.
- The trained ANN achieved 83% sensitivity and 100% specificity.
- The model demonstrated high accuracy in distinguishing normal from keratoconus corneas.
Conclusions:
- Artificial neural networks can accurately classify keratoconus from Orbscan II tests.
- ANNs offer a valuable tool to support clinicians in keratoconus diagnosis.
- This AI approach enhances the diagnostic capabilities for corneal diseases.
