Automatic segmentation in three-dimensional analysis of fibrovascular pigmentepithelial detachment using high-resolution optical coherence tomography
- Christian Ahlers (christian.ahlers{at}meduniwien.ac.at),
- Christian Simader (christian.simader{at}meduniwien.ac.at),
- Wolfgang Geitzenauer (wolfgang.geitzenauer{at}meduniwien.ac.at),
- Geraldine Stock (geraldine.stock{at}meduniwien.ac.at),
- Paul Stetson,
- Shawn Dastmalchi,
- Ursula Schmidt-Erfurth
- Medical University Vienna, Austria
- Medical University Vienna, Austria
- Medical University Vienna, Austria
- Medical University Vienna, Austria
- Carl Zeiss Meditec, United States
- Carl Zeiss Meditec, United States
- Medical University Vienna, Austria
- Published Online First 26 October 2007
Abstract
Background/Aims: A limited number of scans compromises conventional optical coherence tomography (OCT) to track chorioretinal disease in its full extension. Failures of edge detection algorithms falsify the results of retinal mapping even further. High-definition-OCT (HD-OCT) is based on raster scanning and was used to visualize localization and volume of intra- and sub-pigment-epithelial (RPE) changes in fibrovascular pigment epithelial detachments (fPED). Two different scanning patterns were evaluated.
Methods: 22 eyes with fPED were imaged using a frequency-domain, high-speed prototype of the Cirrus™ HD-OCT. Axial resolution was 6 μm; scanning speed was 25 kA-scans/second. Two different scanning patterns covering an area of 6x6mm in the macular retina were compared. Three-dimensional topographic reconstructions and volume calculations were performed using MATLAB™ based automatic segmentation software.
Results: Detailed information about layer-specific distribution of fluid accumulation and volumetric measurements can be obtained for retinal- and sub-RPE volumes. Both raster scans show a high correlation (p<0,01;R2>0,89) of measured values, i.e. PED volume/area, retinal volume and mean retinal thickness. Quality control of the automatic segmentation revealed reasonable results in over 90% of the examinations.
Conclusion: Automatic segmentation allows for detailed quantitative and topographic analysis of the RPE and the overlying retina. In fPED the 128x512 scanning-pattern shows mild advantages when compared to the 256x256-scan. Accompanied with the ability for automatic segmentation, HD-OCT clearly improves the clinical monitoring of chorioretinal disease by adding relevant new parameters. HD-OCT is likely capable to enhance the understanding of pathophysiology and treatment benefit of current anti-CNV strategies in future.
