Purchase this article with an account.
Deborah Goren, John G. Flanagan; Optimization of Threshold Estimation Algorithms for Visual Field Assessment. Invest. Ophthalmol. Vis. Sci. 2011;52(14):5509.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To determine an optimal threshold estimation algorithm for visual field (VF) evaluation using the Heidelberg Edge Perimeter.
Visual field simulations were run using a custom simulator within the Heyex database system used by all Heidelberg Engineering instruments. Previously measured sensitivity values are used as the basis for simulations and "n" iterations can be run from any preselected database. In addition custom VFs can be generated. Each point was tested using a frequency of seeing curve. All responses were based on their position on the frequency of seeing curve. 3 patient types (poor, average and perfect) were simulated using different slopes, and false positive/negative rates. Threshold estimation strategies were tested including basic staircases, adaptive staircases, MOBS and ZEST. Different step sizes were tested along with different rules for completion (e.g. minimum number of reversals versus value within XdB of age pdf). Checks for completion (e.g. retesting any point less than YdB) were also tested. Average number of trials was used to measuretest length. Mean of difference (MoD +/- CLs) was calculated between simulated and original data to measure accuracy of the test.
Primary points were tested with 422 and 421 with a minimum of 3 reversals. MoD values for staircases were similar, but 422 staircases had less trials (Poor patients: 314 vs 326). Secondary points were tested with 42 and 22 staircases for 2 different completion rules (minimum 2 reversals vs. within 3dB of age pdf). Completion requiring 2 reversals had less trials than 3dB age pdf for both 42 and 22 staircases, but larger MoDs. This difference was most pronounced for completion requiring 2 reversals (42:-0.54; 22:-2.4). MoDs were very similar for 42 and 22 staircases with completion of 3dB (-0.31 vs. -0.66), but 42 staircases had less trials (255 vs. 293). Checks for completion increased the number of trials (293 vs. 366), and did not have a large effect on MoD (-0.66 vs. -0.07).
Primary points with 422 staircases with 3 reversals, followed by secondary points with 42 staircases and completion for points within 3dB of age-matched-normals produced the best combination of speed and accuracy.
This PDF is available to Subscribers Only