Automated quantification of drusen and choroid using clinical SD-OCT

Abstract

 

Purpose

 

Drusen and choroidal thinning are known to correlate with age-related macular degeneration (AMD). However, quantification of both features has so far been difficult. We report an automated method to quantify the drusen load and choroidal thickness from clinical spectral-domain optical coherence tomography (SD-OCT).

 

Methods

 

We selected 7 subjects (60-68 yrs) from the Rotterdam Study who had distinct or indistinct soft drusen on fundus photographs. Subjects had undergone macula-centered SD-OCT imaging (Topcon, 512×128×480 voxels, 6.0×6.0×1.7mm³, voxel size of 11.7×46.9×3.5µm³). Enhanced depth imaging was not used. We segmented drusen and choroid separately: For the drusen detection, a graph-search method simultaneously segmented two surfaces, Bruch’s membrane and the surface at the transition of the myoid to the ellipsoid regions of the inner segments (Fig. 1.F). A drusen probability map was defined based on the magnitude of the deviation of local thickness from normal values (Fig. 1.H). The total area of the detected drusen footprints defined the drusen load. For the choroidal segmentation, choroidal vessels were segmented using Hessian object extraction followed by region-growing (Fig. 1.B). The outer boundary of the choroidal vessel-network was modeled as a smooth 3D surface that was fitted to the segmented choroidal vessels (Fig. 1.C). Choroidal thickness map was derived from local distances between the fitted surface and Bruch’s membrane.

 

Results

 

Drusen load quantification and choroidal thickness assessment were quantified successfully in all subjects. Choroidal thickness was 200.0µm on average [95% CI 189.8µm-210.2µm] and drusen load was 0.0107mm² on average [95% CI 0.0085mm²-0.0130mm²] (Fig. 2).

 

Conclusions

 

We have developed an automated method to quantify the presence, distribution and sizes of drusen and determine choroidal thickness from clinical SD-OCT. Our results may be used to start quantitative studies of relationships between drusen and choroidal thinning.

 

 

Figure 1. A) Original SD-OCT B-scan; B) Choroidal Vessel Segmentation performed in 3D; C) Segmenttaion of the outer boundary of choroidal vessel network; D) Thickness Map of choroidal vessel network; E) Original B-scan with drusen (red arrow); F) Sub-retinal layer segmentation; G)Sub-retinal thickness map; H) Drusen detection result.

 

 

Figure 2. Scatter plot of drusen footprint size and choroidal vessel-network thickness.

 

Keywords: 549 image processing • 504 drusen • 452 choroid