May 2006
Volume 47, Issue 13
ARVO Annual Meeting Abstract  |   May 2006
Diagnostic Aid Software for Glaucoma Detection With the Visual Field Test
Author Affiliations & Notes
  • D. Wroblewski
    BioFormatix, Inc., Escondido, CA
  • B.A. Francis
    Department of Ophthalmology, Doheny Eye Institute, Los Angeles, CA
  • D.S. Minckler
    Department of Ophthalmology, Doheny Eye Institute, Los Angeles, CA
  • P. Quiros
    Department of Ophthalmology, Doheny Eye Institute, Los Angeles, CA
  • V. Chopra
    Department of Ophthalmology, Doheny Eye Institute, Los Angeles, CA
  • R.K. Massengill
    Mednovus, Leucadia, CA
  • Footnotes
    Commercial Relationships  D. Wroblewski, BioFormatix, Inc., E; Developer, P; B.A. Francis, BioFormatix, Inc., C; D.S. Minckler, BioFormatix, Inc., C; P. Quiros, BioFormatix, Inc., C; V. Chopra, BioFormatix, Inc., C; R.K. Massengill, BioFormatix, Inc., C.
  • Footnotes
    Support  NIH Grant 2 R44 EY014077
Investigative Ophthalmology & Visual Science May 2006, Vol.47, 3976. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      D. Wroblewski, B.A. Francis, D.S. Minckler, P. Quiros, V. Chopra, R.K. Massengill; Diagnostic Aid Software for Glaucoma Detection With the Visual Field Test . Invest. Ophthalmol. Vis. Sci. 2006;47(13):3976.

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

Purpose: : To evaluate the applicability of novel pre–processing approaches and pattern recognition methods for automatic interpretation of the visual field test and glaucoma detection.

Methods: : The visual field (VF) test data were obtained with the standard automated perimetry device (HFA II using 24–2 algorithm). The ancillary data included treated and untreated intra–ocular pressure (IOP), cup–to–disk ratio, patient’s age, race, sex, and family history risk factor. Classification of visual fields and the ophthalmic diagnosis (based on all relevant data) was provided by experts. 216 cases including normal, glaucoma suspect, and glaucoma patients were analyzed. Two distinct classification schemes were studied: (1) classification of visual field test data into normal, consistent with glaucoma, and glaucomatous, (2) classification into glaucoma diagnostic classes of normal, glaucoma suspect and glaucoma. For the present dataset, the analysis of VF data only results in under diagnosis of glaucoma. The VF data were pre–processed to produce parameters related to sensitivity distribution, which were subsequently used by non–parametric classification algorithms (radial basis function and multilayer perceptron neural networks, and support vector machines). Classification accuracy (with respect to the ‘ground truth’ expert evaluation) was determined using advanced numerical validation methods (random cross–validation and bootstrap).

Results: : The study indicated the feasibility of accurate detection of glaucoma and suspect glaucoma from the visual field data, only. With most of the applied methods, the specificity (rate of correctly classified normal eyes) was over 99%, and the rate of suspect glaucoma and glaucoma detection was in the 93% and 98% range, respectively. The numerical models were able to detect cases of suspect glaucoma, which are often associated with visual fields that are considered normal.

Conclusions: : The results indicate superior performance of newly developed modeling approaches compared to those previously reported in the literature. The study is presently being expanded to include a larger database of over 2000 cases of different stages of glaucoma progression, neurological effects, and artifactual data, with the aim of providing an automated screening tool for early detection of glaucoma.

Keywords: visual fields 

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.