Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Current keratoconus detection methods compared with a neural network approach

M K Smolek1, S D Klyce

  • 1LSU Eye Center, Louisiana State University Medical Center, New Orleans 70112, USA.

Investigative Ophthalmology & Visual Science
|October 31, 1997
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Developments in corneal topographic analysis following contact lens wear and refractive surgery.

Contact lens & anterior eye : the journal of the British Contact Lens Association·2005
Same author

Contribution of the ocular surface to visual optics.

Experimental eye research·2004
Same author

Screening of prior refractive surgery by a wavelet-based neural network.

Journal of cataract and refractive surgery·2001
Same author

Effect of larger ablation zone and transition zone on corneal optical aberrations after photorefractive keratectomy.

Archives of ophthalmology (Chicago, Ill. : 1960)·2001
Same author

Gender- and age-related differences in corneal topography.

Cornea·2001
Same author

Corneal topography and the new wave.

Cornea·2000

Neural networks accurately detect keratoconus (KC) and keratoconus suspects (KCS), outperforming traditional methods in accuracy and specificity for corneal topography analysis.

Area of Science:

  • Ophthalmology
  • Artificial Intelligence
  • Medical Imaging

Background:

  • Keratoconus (KC) is a progressive corneal ectasia.
  • Accurate detection and grading of KC are crucial for timely intervention.
  • Existing videokeratographic methods have limitations in distinguishing KC from similar conditions.

Purpose of the Study:

  • To compare the diagnostic performance of four videokeratographic methods against a novel neural network approach for keratoconus detection.
  • To evaluate the ability of neural networks to screen for KC and keratoconus suspects (KCS) and grade KC severity.

Main Methods:

  • A classification neural network was developed to identify KC, KCS, or other corneal conditions.
  • A separate severity network graded the degree of conelike topography.

Related Experiment Videos

  • Three hundred TMS-1 examinations were used for training and testing the networks.
  • Ten topographic indices served as inputs, with nine output categories including normal, KC stages (KC1-KC3), and KCS.

    • Main Results:

    • The classification network achieved 100% accuracy, specificity, and sensitivity on the test set.
    • The severity network demonstrated strong correlation with the keratoconus prediction index (R = 0.892, P < 0.0001).
    • The neural network approach showed significantly better accuracy and specificity compared to Klyce/Maeda keratoconus index (KCI) and Rabinowitz (K & I-S) tests, and average simulated keratometry (Sim K).

    Conclusions:

    • Neural networks effectively differentiate keratoconus (KC) from keratoconus suspects (KCS) and other mimicking topographies.
    • The network approach matched the sensitivity of current tests while surpassing them in accuracy and specificity.
    • This AI-driven method offers a promising advancement in objective keratoconus screening and grading.