Purchase this article with an account.
W.H. Swanson, M.W. Dul, J.H. Sohn, F. Pan; Use of Linear Units for Comparing Perimetric and Morphologic Measures of Glaucomatous Damage . Invest. Ophthalmol. Vis. Sci. 2003;44(13):57.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose: Perimetric sensitivity in glaucomatous eyes has a curvilinear relation to neuroretinal rim area and to ganglion cell loss. An empirical study found that use of a linear scale for perimetric sensitivity can make the relation linear (Garway-Heath et al., IOVS 43:2213). We used a quantitative neural model of perimetric sensitivity to analyze their finding, and applied their analysis to clinical data on individual sectors of the neuroretinal rim. Methods: The model has two stages: the first computes responses of a mosaic of ganglion cells, the second computes weighted sums of ganglion cell outputs as responses of cortical spatial filters. Perimetric defects were simulated by random removal of ganglion cells from the mosaic in first stage. The application was to data from 20 eyes of 20 patients with glaucoma, selected from the Glaucoma Institute database using the following criteria: diagnosis of glaucoma with no concomitant disorders that affect vision, good optic nerve image and disc outline on the Heidelberg Retinal Tomograph (HRT), reliable 24-2 visual field within 6 months of the HRT image. To ensure uniform distribution of the independent variable, 5 eyes were selected for each of 4 groups defined by global HRT neuroretinal rim area. Each perimetric location was associated with one of six sectors of the optic disc map of Garway-Heath et al. (Ophthalmology 107:1809). For each eye, the average perimetric sensitivity corresponding to a given sector was computed in linear units. Results: For the model, depth of defect was linearly related to amount of ganglion cell loss, due to the fact that the second-stage cortical populations sum the responses of many ganglion cells. For the clinical data, perimetric sensitivity in linear units was highly correlated with rim area for the temporal half of the nerve(r=0.78, p less than 0.0001) but not the nasal half (r=0.41, p greater than 0.07). Perimetric sensitivity decreased by 58% when area of corresponding temporal rim decreased by 50%. Similar results were found when the six sectors were evaluated individually: r greater than 0.59, p less than 0.01 for temporal sectors, r less than 0.36, p greater than 0.07 for nasal sectors. Conclusions: A quantitative model provides support for the assertion that a linear scale for perimetric sensitivity is appropriate for correlating structure and function. Application of this approach to clinical data found that perimetric sensitivity expressed in linear units was highly correlated with corresponding temporal neuroretinal rim area, with approximately equal percentage loss for both measures.
This PDF is available to Subscribers Only