April 2014
Volume 55, Issue 13
Free
ARVO Annual Meeting Abstract  |   April 2014
A New Visual Field Testing Algorithm that Better Detects Glaucomatous Progression
Author Affiliations & Notes
  • Andrew Turpin
    Computing and Information Systems, University of Melbourne, Melbourne, VIC, Australia
  • Luke Xiang-Yu Chong
    Computing and Information Systems, University of Melbourne, Melbourne, VIC, Australia
    Department of Optometry and Vision Sciences, University of Melbourne, Melbourne, VIC, Australia
  • Allison M McKendrick
    Department of Optometry and Vision Sciences, University of Melbourne, Melbourne, VIC, Australia
  • Footnotes
    Commercial Relationships Andrew Turpin, Haag-Streit (F), Heidelberg Engineering (F); Luke Chong, None; Allison McKendrick, Haag-Streit (F), Heidelberg Engineering (F)
  • Footnotes
    Support None
Investigative Ophthalmology & Visual Science April 2014, Vol.55, 5641. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Andrew Turpin, Luke Xiang-Yu Chong, Allison M McKendrick; A New Visual Field Testing Algorithm that Better Detects Glaucomatous Progression. Invest. Ophthalmol. Vis. Sci. 2014;55(13):5641.

      Download citation file:


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

      ×
  • Supplements
Abstract
 
Purpose
 

The GOANNA algorithm (IOVS 54:3943) uses gradients to choose test locations such that visual fields are more densely sampled around edges of visual field loss. Thus, we hypothesize that GOANNA should be better at detecting glaucomatous progression than methods using a fixed 24-2 pattern.

 
Methods
 

GOANNA tests predefined seed locations using a partial Zippy Estimation by Sequential Testing (ZEST) procedure, and then chooses subsequent test locations from a 3x3 degree grid based on the maximum difference between already tested locations divided by distance apart. New locations have their prior set using natural neighbor interpolation, and locations that are never tested are finally assigned thresholds using the same. As no progressing dataset on a 3x3 degree grid exists, we derived a dataset using the method of Spry et al (IOVS 41) from measured visual fields for 23 people with glaucoma obtained on a 3x3 degree grid (central 27 degrees of visual field, total of 150 locations). A sequence was labeled as progressing if 3 or more locations triggered the Two-Omitting Point-wise Linear Regression criteria of Gardiner et al (IOVS 43). A stable dataset of 230 sequences was similarly created. Using simulation, three procedures were compared: GOANNA; ZEST for 24-2 locations (Turpin et al IOVS 44); and the ZEST procedure on 24-2 locations with the other 98 locations in the 3x3 deg grid found using natural neighbor interpolation. Specificity and sensitivity was varied by altering the p-value used to call progression for the point-wise linear regression. Receiver operating characteristic (ROC) curves were plotted with greater area under the ROC curve (AUC) being considered better performance. Both reliable patients (0% false responses) and false positive patients were simulated (15% false positive rate, 3% false negative rate).

 
Results
 

Figure 1 shows that GOANNA outperformed ZEST for reliable patients, and was about the same for false positive patients.

 
Conclusions
 

The 24-2 test pattern samples a minority of visual space, hence will miss some visual field defects and their early spatial expansion. GOANNA is designed to enable customized spatial testing, thus detects glaucomatous progression earlier than ZEST using the 24-2 pattern while using the same number of presentations.

 
 
Symbols show area under the ROC curve for calling progression at each 6 month visit using each algorithm. Error bars show a 95% confidence interval.
 
Symbols show area under the ROC curve for calling progression at each 6 month visit using each algorithm. Error bars show a 95% confidence interval.
 
Keywords: 758 visual fields • 642 perimetry  
×
×

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.

×