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

Visual field analysis using artificial neural networks

S E Spenceley1, D B Henson, D R Bull

  • 1Department of Optometry and Vision Sciences, University of Wales College of Cardiff, UK.

Ophthalmic & Physiological Optics : the Journal of the British College of Ophthalmic Opticians (Optometrists)
|July 1, 1994
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

Noise creates polarization artefacts.

Bioinspiration & biomimetics·2017
Same author

New superior-inferior visual field asymmetry indices for detecting POAG and their agreement with the glaucoma hemifield test.

Eye (London, England)·2015
Same author

Individualised patient care as an adjunct to standard care for promoting adherence to ocular hypotensive therapy: an exploratory randomised controlled trial.

Eye (London, England)·2011
Same author

Pascal panretinal laser ablation and regression analysis in proliferative diabetic retinopathy: Manchester Pascal Study Report 4.

Eye (London, England)·2011
Same author

High-resolution hyperspectral imaging of the retina with a modified fundus camera.

Journal francais d'ophtalmologie·2010
Same author

Preliminary survey of educational support for patients prescribed ocular hypotensive therapy.

Eye (London, England)·2010
Same journal

Mesopic Illumination in Natural Environments: Implications for Myopia Research.

Ophthalmic & physiological optics : the journal of the British College of Ophthalmic Opticians (Optometrists)·2026
Same journal

Axial Length Growth Reference Curves and LMS Parameters for Japanese Children and Adolescents Aged 4-20 Years: The TMM BirThree Cohort Study.

Ophthalmic & physiological optics : the journal of the British College of Ophthalmic Opticians (Optometrists)·2026
Same journal

Effect of Switching Myopia Control Strategies on Axial Elongation in Children with Poor Response to Highly Aspherical Lenslet Therapy.

Ophthalmic & physiological optics : the journal of the British College of Ophthalmic Opticians (Optometrists)·2026
Same journal

Quantifying the Role of Longitudinal Chromatic Aberration and Age in Night Vision Disturbances.

Ophthalmic & physiological optics : the journal of the British College of Ophthalmic Opticians (Optometrists)·2026
Same journal

Impaired Rod-Mediated Vision is the Functional Hallmark of Ageing and Early and Intermediate Age-Related Macular Degeneration.

Ophthalmic & physiological optics : the journal of the British College of Ophthalmic Opticians (Optometrists)·2026
Same journal

Analytical Performance, Spatial Dynamics and Clinically Meaningful Change Thresholds for Automated Non-invasive Tear Film Assessment Using the Oculus Keratograph 5M.

Ophthalmic & physiological optics : the journal of the British College of Ophthalmic Opticians (Optometrists)·2026
See all related articles

Optimizing artificial neural networks (ANNs) for glaucoma detection requires large, balanced training datasets. The ideal set includes diverse, early-stage glaucomatous defects across all spatial locations for improved sensitivity and specificity.

Area of Science:

  • Ophthalmology
  • Artificial Intelligence
  • Medical Image Analysis

Background:

  • Artificial neural networks (ANNs) show promise in visual field classification.
  • Previous studies have not fully explored training set impact on ANNs for visual field analysis.
  • Differentiating normal from glaucomatous visual fields is a critical diagnostic challenge.

Purpose of the Study:

  • To investigate the effect of training set characteristics on the performance of ANNs in classifying visual fields.
  • To identify optimal training set properties for accurate glaucoma detection using ANNs.

Main Methods:

  • Utilized a multilayer neural network with 132 input, 20 hidden, and 2 output nodes, trained via error backpropagation.
  • Analyzed visual field data from a Henson CFS2000 perimeter.

Related Experiment Videos

  • Evaluated training set properties: size, class balance (normal vs. glaucoma), extent of field loss, and spatial defect distribution.
  • Measured network performance using sensitivity and specificity.

    • Main Results:

    • Larger training sets improved sensitivity without compromising specificity, and reduced performance variability.
    • Class-biased sets (more glaucoma or normal examples) skewed sensitivity and specificity accordingly.
    • The spatial distribution of defects in the training set significantly impacted the network's ability to generalize.
    • Training sets lacking specific defect locations hindered performance.

    Conclusions:

    • Large, balanced training sets with diverse examples of early glaucomatous defects across all spatial locations are crucial for optimal ANN performance in visual field classification.
    • The composition and characteristics of the training data are as important as the network architecture itself for accurate diagnostic tools.