Purchase this article with an account.
P. M. Brennen, R. A. Bilonick, T. R. Friberg; Drusen Area Calculations on Fundus and Corresponding En-face OCT Images: A Comparison. Invest. Ophthalmol. Vis. Sci. 2010;51(13):303.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Semi-automated methods are available for the calculation of drusen area based on fundus images. Recently, several groups have used 3D SDOCT data for drusen volume and area calculations. We compared standard fundus photographs (FP) with en-face OCT images of drusen using a semi-automated drusen detection technique.
We retrospectively identified patients with dry macular degeneration who had 3D SDOCT data (Cirrus HD-OCT, Carl Zeiss Meditec, Dublin, CA) and fundus photography on the same day. An en-face OCT image was generated to represent drusen distribution in the macula by sampling along the Z axis above the RPE using the Advanced Visualization tool. Inclusion of low reflectivity outer plexiform layer A scan data provided contrast to high reflectivity drusen and RPE data for optimal images. OCT images and FPs were analyzed with the Drusen Analyzer software (F.O.R.T.H., Crete, Greece) to obtain the drusen area (mm2) in a 3 mm diameter circle centered at the fovea. Two measurements were performed for each source data type.
En-face OCT and fundus image pairs were identified in and analyzed in ten eyes of 7 patients. Plotting of the differences versus the averages is presented in the Figure. When comparing OCT and FPs, a scale bias of 1.54 was demonstrated. Area measurements from FPs were larger than those obtained from OCT images. When comparing imprecision, the relative imprecision of fundus photography as estimated by the method of maximum likelihood was much less, only 7% of the OCT imprecision.
Our results from the sample eyes suggest poor agreement between OCT and fundus photographic image based drusen area assements. The higher relative imprecision of the OCT method suggests that further optimization is needed. Precise explanation of the scale bias is difficult as the absolute accuracy of the differing approaches had yet to be determined. Considering the presence of scale bias and degree of imprecision, it is likely that drusen volume calculations predicated on detection of drusen area based on FPs will differ from drusen volume calculated from OCT data alone.
This PDF is available to Subscribers Only