FDT Visual Field Simulation Results using Barramundi

A. Turpin, A.M. McKendrick, C.A. Johnson, and A.J. Vingrys.

This site accompanies the paper "Development of efficient threshold strategies for Frequency Doubling Technology perimetry using computer simulation". Invest Ophthalmol Vis Sci 43, Jan 2002. Pages 308-313.

Introduction

This www page details results of experiments simulating the performance of the MOBS and ZEST visual field algorithms in normal and glaucomatous patients for Frequency Doubling Technology Perimetry. It is a detailed list of our simulation of 1080 test procedures on 6 groups of patients, and 3 artificial patients. A first time reader is referred to a much more accessible summary of the main results that have been published as...

The simulations were carried out using Barramundi, software written in Java by Andrew Turpin while working as a postdoc at Discoveries In Sight, Portland, OR.

Barramundi has also been used to simulate SWAP (blue-on-yellow) and SAP (white-on-white) perimetry, but those results are not reported here.

Methods

Barramundi accepts three inputs:

1. a list of patient thresholds, in this case 21 dB values;
2. a test procedure coded as a Java class; and
3. a description of the patients behaviour as a false neagative rate, a false positive rate, and the distribution of other errors.

A more detailed overview of Barramundi can be found by clicking on this Barramundi, link.

The following tables decribe the patient sets and test procedures used in this study.

Patient sets

Test procedures

A ZEST procedure is described by four components: a letter Z, a pdf code from Table 3, a mlf code from Table 4, and a stopping condition from Table 5. For example, the code Z-n-99_3-n3 describes the ZEST procedure using the normal pdf, a steep mlf with 99% asymptotes, and stopping after 3 presentations. Given that there are 15 possible pdfs, 12 possible mls, and 6 possible stopping conditions, there are a total 0f 15*12*6 = 1080 ZEST procedures availalble for testing.

Results

The results reported are

1. the number of presentations required by a method on a patient set,
2. the difference between the input threshold and determined threshold, and
3. the absolute value of the difference between the input threshold and determined threshold.

Raw data in space delimited format for the following 9 tables can be downloaded by clicking on the link at the beginning of this sentence. The download file is a zip archive of 9 files, one per table.

Table 6: All test procedures on perfect normals (patient set 1)

• Sorted by procedure
• Sorted by mean number of presentations
• Sorted by mean absolute accuracy

Table 7: All test procedures on perfect glaucomas (patient set 4)

• Sorted by procedure
• Sorted by mean number of presentations
• Sorted by mean absolute accuracy

Table 8: All test procedures on typical normals (patient set 2)

• Sorted by procedure
• Sorted by mean number of presentations
• Sorted by mean absolute accuracy

Table 9: All test procedures on typical glaucomas (patient set 5)

• Sorted by procedure
• Sorted by mean number of presentations
• Sorted by mean absolute accuracy

Table 10: All test procedures on bad normals (patient set 3)

• Sorted by procedure
• Sorted by mean number of presentations
• Sorted by mean absolute accuracy

Table 11: All test procedures on bad glaucomas (patient set 6)

• Sorted by procedure
• Sorted by mean number of presentations
• Sorted by mean absolute accuracy

Table 12: All test procedures on artifical patient sets (7, 8 and 9)

• No error (Patient set 7)
• Typical error (Patient set 8)
• Bad error (Patient set 9)

Created Aug 30 2001 by Andrew Turpin

Last updated Mon Sep 3 12:46:02 WST 2001