Purchase this article with an account.
K.A. Muldrew, R.C. McGivern, M.R. Stevenson, U. Chakravarthy; Parametric Analysis of Fluorescein Angiograms in Choroidal Neovascularisation Secondary to Age–related Macular Degeneration . Invest. Ophthalmol. Vis. Sci. 2004;45(13):2958.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose: To develop computerised methods to quantify area and filling characteristics of choroidal neovascularisation (CNV) using fluorescein angiography (FA). Methods: Seventy–nine digitally acquired angiograms from patients who participated in the subfoveal radiotherapy study with CNV secondary to AMD were analysed. Two methods were used to delineate the area of abnormal fluorescence in each frame of each angiogram: manual tracing of outlines of the regions of interest and a computerised thresholding algorithm. Analysis was undertaken by fitting the data with an empirical function and extracting values representing the maximum area of hyperfluorecence (MAHF) and time to 63% (τc) of MAHF. These parameters were examined for relationships with clinical measures of vision (distance and near visual acuity and contrast sensitivity) and CNV subtype classification (predominantly classic or minimally classic) using logistic regression techniques. Results:Forward–Wald binary logistic regression in which the dependent variable was CNV subtype with the following covariates; MAHF from both manual and thresholded outlining, τc from both methods and three clinical measures of vision. The only factors significant in the probability of allocation to the correct CNV subtype were MAHF from manual outlining and logNVA (p=0.001 and p=0.002 respectively). The received operator curve (ROC) provided a predictive probability of 0.79 for MAHF from manual outlining and logNVA. Conclusions: The present study has demonstrated that computer assisted techniques are useful in the parametric analysis and interpretation of CNV subtype in FA. Notably MAHF derived from manual outlining and logNVA yielded most informative data to permit CNV subtype classification. Worse near visual acuity and small lesions were predictive of a predominantly classic CNV subtype.
This PDF is available to Subscribers Only