March 2012
Volume 53, Issue 14
Free
ARVO Annual Meeting Abstract  |   March 2012
Improved Prediction Of Function From Structure In Glaucoma: Handling Visual Field Variability
Author Affiliations & Notes
  • Haogang Zhu
    Optometry and Visual Science, City University London, London, United Kingdom
    NIHR Biomedical Research Centre for Ophthalmology, Moorfields Eye Hospital NHS Foundation Trust and UCL Institute of Ophthalmology, London, United Kingdom
  • David P. Crabb
    Optometry and Visual Science, City University London, London, United Kingdom
  • David F. Garway-Heath
    NIHR Biomedical Research Centre for Ophthalmology, Moorfields Eye Hospital NHS Foundation Trust and UCL Institute of Ophthalmology, London, United Kingdom
  • Footnotes
    Commercial Relationships  Haogang Zhu, None; David P. Crabb, None; David F. Garway-Heath, Carl Zeiss Meditec (F), Heidelberg Engineering (F), OptoVue (F)
  • Footnotes
    Support  Haogang Zhu is supported by research fellowship from National Institute for Health Research, UK
Investigative Ophthalmology & Visual Science March 2012, Vol.53, 701. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Haogang Zhu, David P. Crabb, David F. Garway-Heath; Improved Prediction Of Function From Structure In Glaucoma: Handling Visual Field Variability. Invest. Ophthalmol. Vis. Sci. 2012;53(14):701.

      Download citation file:


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

      ×
  • Supplements
Abstract
 
Purpose:
 

A Bayesian radial basis function (BRBF) allows visual field (VF) sensitivity to be predicted from retinal nerve fibre layer (RNFL) measurements (Zhu et al, IOVS, 51: 5657-66, 2010). A VF may also be predicted from other VFs from the same individual. The BRBF did not account for the increased variability as VF sensitivity declines and resulted in biased estimates of function, especially in regions of better sensitivity in glaucomatous eyes. This work sets out to improve the BRBF by accounting for the variable variability at different VF sensitivities.

 
Methods:
 

A new BRBF (vBRBF) was developed to the predict VF sensitivity, and the confidence of the prediction, from RNFL images. The model was trained on data from 119 patients and 110 visually healthy subjects with GDxVCC RNFL thickness (RNFLT) and standard automated perimetry (SAP) measurements and was tested on independent data from 76 patients and 230 healthy subjects (Zhu et al, IOVS, 2010). The point-wise VF sensitivities were predicted from the RNFL map, and the prediction error (mean absolute difference [MAD]), was compared for vBRBF and BRBF. As an independent validation, the vBRBF was used to predict the VF sensitivity from a VF data set acquired on 50 patients each with 5 repeat VFs within a short period of time (Zhu et al, Arch Ophthalmol, 129: 1167-74, 2011). The confidence in the prediction (range between 5% and 95% of possible values) from a RNFLT image and from another VF was then compared.

 
Results:
 

The MAD (mean±std) of vBRBF is 2.5dB ± 3.1dB compared with that of BRBF 2.8 dB ± 3.8 dB. The improvement is significant (p<0.001, paired t-test) with VF sensitivity ≥20dB (Figure (a)). The confidence in the prediction (Figure (b)) from RNFLT images accords well (p>0.1, paired t-test) with that from VFs above 15dB and this confidence is weaker as sensitivity decreases.

 
Conclusions:
 

vBRBF was able to correct the prediction bias present in BRBF. A VF is predictable from GDxVCC RNFLT measurement but the confidence of the prediction worsens as sensitivity declines. Combining structure and function measurement may integrate the usefulness of imaging techniques and SAP across the range of disease and may improve the monitoring of glaucomatous change.  

 
Keywords: nerve fiber layer • visual fields • computational modeling 
×
×

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.

×