Purchase this article with an account.
P. Updike, A. Muldrew, S.R. Sadda, A.C. Walsh, U. Chakravarthy; Quantitative Fluorescein Angiography of Choroidal Neovascularization . Invest. Ophthalmol. Vis. Sci. 2006;47(13):4494.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To compare measurements from newly developed computer software with traditional grading methods in the analysis of fluorescein angiograms from patients with choroidal neovascularization (CNV) due to age–related macular degeneration and to examine the relevance of the parameters generated by the former by testing their relationships with distance visual acuity.
Custom quantitative fluorescein analysis (QFA) software was used to analyze sixty one angiograms from patients with CNV. Distance visual acuity for the corresponding visit was available for each angiogram. QFA generated 3 mathematical parameters (pixel area [PA], integrated intensity [II], positive fluorescence [PF]) for each of the following lesion components (classic CNV, occult CNV, blocked fluorescence due to blood, blocked fluorescence other and leakage). The background and optic nerve fluorescence were also divided into two parameters, II and PF, to produce 4 exposure–corrected quotients, IIb, IIo, PFb and PFo. The angiograms were also analyzed by trained graders and assigned to one of 3 subgroups based on the proportion of classic CNV in the lesion. The area of classic CNV and lesion area were estimated using MPS disc area circles to yield graded categorical values.
Agreement between QFA and traditional grading was high for lesion subtype assessment (kappa 0.66). Regression analyses were performed with DVA as the dependent variable and PA, II, IIQb, IIQo, PFb and PFo for each lesion component as the independent variables. In every case, the PFb quotient exhibited the most significant relationship with the best model (adjusted r2 0.27) seen with components of classic CNV. The parameter estimates showed that a change of 30 on the PFb for classic CNV, resulted in a loss of 20 letters (4 lines). None of the models using parameters generated by traditional grading showed statistically significant relationships with distance acuity.
Computerized analysis was equivalent to traditional grading methods in assigning lesion subtype. Quantitative parameters describing the filling characteristics of classic CNV exhibited stronger relationships with visual acuity than graded categorical estimates of either lesion size or classic CNV.
This PDF is available to Subscribers Only