Abstract

P 252

Performance of drusen detection by current spectral-domain optical coherence tomography

Ferdinand Georg Schlanitz, Christian Ahlers, Christopher Schütze, Géraldine Stock, Isabelle Golbaz, Stefan Sacu, Ursula Schmidt-Erfurth
Klinik für Augenheilkunde und Optometrie, AKH, Medizinische Universität Wien, Wien, Österreich

Objective
Aim of this study was to evaluate the performance of segmentation procedures integrated in current spectral-domain optical coherence tomography (SD-OCT) devices (Cirrus, Carl Zeiss Meditec; 3D-OCT-1000, Topcon and HRA Spectralis, Heidelberg Engineering) to identify drusen in patients with non-atrophic and non-neovascular AMD.
Methods
17 eyes of 13 patients with drusen of AREDS III/IV classification were examined with three clinical SD-OCT devices (Cirrus, 3D-OCT-1000 and HRA Spectralis). After performing standard automatic segmentation procedures, every druse in an individual B-scan was documented and compared to the automatic segmentation result. Errors in the segmentation were classified as negligible, medium and severe. Relations between the size of the druse and its appearance in the segmentation algorithm were evaluated. Drusen identified by an expert grader (EG) were plotted using a custom made Matlab program to compare identification and localization of a druse in the specific OCT device.
Results
The interpolated segmentation of the retinal pigment epithelium (RPE) of Cirrus made significantly (p <0,001) less errors in detecting drusen than 3D-OCT1000. The Cirrus 200x200 Scan was able to detect 33% of the identified drusen with negligible error (22% in the 512x128 Scan); 3D-OCT1000 detected <7,6% of the drusen with negligible errors. Spectralis did not offer a true RPE-segmentation. The substituted identification of Bruchs membrane was not able to identify drusen in the macula. The mean extend of a successfully detected druse was 270 µm (SD 121) in the Cirrus 200x200 Scan (343 µm (SD 140) in the 512x128 Scan). Counting the plotted drusen identified by an EG resulted in over 200% more drusen than seen on the standard fundus photography in some cases.
Conclusions
Although current SD-OCT devices showed significant potential in detecting drusen, the current segmentation algorithms failed to detect all drusen present in the individual B-Scans reliably in this trial. Automatic segmentation procedures should be further developed to obtain a better performance in future. The significantly higher amount of drusen detected by the expert examiner in SD-OCT Scans compared to fundus photography shows the potential of SD-OCT in the diagnostic of the early age-related macular disease.

Copyright: 2009 Porstmann Kongresse GmbH. All rights reserved. Impressum, Disclaimer.