A method for automated choroidal-scleral interface segmentation in optical coherence tomography

Abstract

Purpose : Reliable choroidal layer analysis has become an important diagnostic tool for retinal diseases. This abstract proposes a choroidal-scleral interface segmentation method in swept-source optical coherence tomography (SS-OCT) volumes based on an iterative graph-based algorithm.

Methods : A relatively fast algorithm was developed to segment the choroidal-scleral interface. The segmenetation starts from a B-scan with high contrast around choroidal-scleral boundary. Then the segmentation is propagated to the entire volume data. The algorithm uses the intensity as well as gradient images as inputs to a graph-based method to segment each B-scan in a region of interest. Performance of the algorithm is evaluated using 49 normal SS-OCT volume data of 500x500 A-scans over 12x12 mm acquired using PLEX^® Elite 9000 SS-OCT (ZEISS, Dublin, CA). The choroidal thickness maps using manual and automated segmentation were generated, defined as the distance between a fitted RPE baseline and the choroidal-scleral interface. The performance of the algorithm is reported using regression and Bland Altman analyses for each sector of the ETDRS grid (Figure 1). The ETDRS grid consists of three concentric circles of 2, 4, and 6 mm radius centered on the fovea.

Results : Figure 1 shows an example of the choroidal thickness map generated by manual and automated segmentation, and corresponding structural choroidal vasculature map. Table 1 shows the information extracted from regression and Bland Altman analyses for each sector of the ETDRS grid. The regression and Bland-Altman analyses for all sectors demonstrate strong correlation and good agreement between the manual and automated method. The average processing time is less than 4 seconds using Intel i7 CPU, 2.7 GHz with 32 GB memory.

Conclusions : We proposed a new choroidal-scleral interface segmentation. Overall, the choroidal thickness maps generated by automated and manual segmentation have a strong correlation and good agreement. Automated segmentation may be a valuable diagnostic tool for retinal diseases.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.

 

Figure 1: Choroidal thickness map (in microns) of a right eye, generated using manual (left) and automated (middle) segmentation, with overlaid ETDRS grid centered on the fovea, and the structural choroidal vasculature map (right)

 

Table 1: Information extracted from regression and Bland Altman analyses of average choroidal thicknesses for all 9 sectors of the EDTRS grid