Purchase this article with an account.
David W. Richards, Samantha Roland, Gretta Fridman, Arthur Snider; Noise-Corrected Maximum-Likelihood Analysis of Global Visual Field Progression. Invest. Ophthalmol. Vis. Sci. 2011;52(14):4155.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To develop a mathematical measure of global visual field change while accounting for all quantifiable effects that interfere with estimation of this change.
We retrospectively analyzed 92 Zeiss-Humphrey (ZH) 24-2 Sita-Standard visual fields (VFs) of 10 eyes of 10 glaucoma patients. There were 7 to11 serial VFs per eye extending over 4 to 8 years. All eyes had acuities of 20/40 or better and combined false positive + false negative rates of < 20%. Mean Defect (MD) ranged from -19 to 0, and Pattern Standard Deviation (PSD) from 2.17 to 15.63. Raw sensitivity (S) and Total Deviation (TD) maps were used in the analysis. A histogram of TD revealed a bimodal distribution, whose low-end secondary peak was determined to be due to non-thresholded (NT) points ( S = "<0") . This is one source of "noise". A second type of noise was the non-Gaussian distribution of TD, after elimination of NT points. A third source was fluctuation over time at individual TD locations ("loci"). Using transformation of variable, elimination of NT points, and calculation of pointwise time-variability of transformed TD at each locus, we calculated the Maximum-Likelihood (ML) global rate of change (R) of TD for each eye, with corresponding 1-sigma uncertainty (sigR). R and sigR were compared with the corresponding parameters (RZH and sigRZH) derived using MD and PSD from ZH as input.
R and RZH were not significantly different (p = 0.16, Mann-Whitney U Test ), but sigR (mean 0.222 db/year) was significantly smaller than sigRZH (mean 1.33 db/year) (p <0.001, Mann-Whitney U Test ).
There is a clinical need for a sensitive measure of global visual field change (Caprioli J, Am J of Ophthalmology 2008; 145; 191-192; Johnson C, Glaucoma Today 2010; 8; 34 ). The standard ZH parameters MD and VFI (Visual Field Index) include averages over all loci within a VF and are subject to contamination by inclusion of NT points, as well as by two other types of noise. Correction for all three types of noise and ML analysis improve the statistical accuracy of measurement of global VF change.
This PDF is available to Subscribers Only