Purchase this article with an account.
J. F. Jennings, D. W. Richards; An Extended Study of Linear Spatial Averaging of Automated Visual Fields in the Context of the Structure-Function Relationship With OCT. Invest. Ophthalmol. Vis. Sci. 2009;50(13):3515.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
In a previous limited study (submitted to ARVO 2009), we showed that linear spatial averaging of automated visual fields reduces scatter of total deviation (TD) in terms of structure-function correlation with OCT. We now present a larger study that confirms the early results and also shows a similar behavior of pattern deviation (PD).
We retrospectively analyzed 136 hemifields of 68 eyes of 56 patients to compare the Zeiss-Humphrey (ZH) TD and PD values with RNFL thickness, obtained using the Zeiss Stratus OCT System with the 4.0.1 software package and the Fast RNFL Program 3.4. Exclusion criteria for ZH visual fields were a fixation loss rate of more than 15%, a combined false positive and false negative rate of over 20%, or a visual acuity less than 20/50. Inclusion criterion for OCT was at least 2 red sectors in one hemifield, a red sector indicating RNFL thickness in the first percentile of age-matched normals. Exclusion criteria for OCT were a signal strength of less than 5 or obvious error in computer identification of RNFL boundaries. Hemifield spatial averages of TD and PD were calculated in the db and linear domains. OCT values for corresponding hemifields were calculated as the average of the upper or lower 5 sectors of OCT data. The linear hemifield averages of TD and PD were converted back to db, and straight-line least-square fits were computed for Case 1: TD (in db) vs. OCT, Case 2: TD (linear) vs. OCT, Case 3: PD (in db) vs. OCT, and Case 4: PD (linear) vs. OCT.
RMS residuals were 7.27 db for Case 1 and 4.10 db for Case 2. RMS residuals were 5.83 db for Case 3 and 2.70 db for Case 4.Statistical significance of the difference between Case 1 and Case 2 was p < 0.001 by both F-test and Kolmogorov-Smirnov test. The same significance, by both tests, was found for the pair Case 3 and Case 4.
In the HVF/OCT structure-function relationship, scatter is significantly reduced for both TD and PD by computing spatial averages in the linear domain. Linear spatial averaging may be preferable to db averaging in all applications of HVF’s.
This PDF is available to Subscribers Only