To develop an automatic method for detecting Pigment epithelial detachment (PED) segmentation instead of traditional manual time consuming and labor intensive segmentation for serous PED objects, and quantifying its volume, size and total number in 2D and 3D spectral domain optical coherence tomography (SD-OCT) images which are important markers of AMD disease severity, progression, and vision loss.

PED detection and volume segmentation was performed using both automated algorithm and manual segmentation. The proposed framework for automatic segmentation consists of four main image processing steps: 1) Preprocessing, to remove possible highly-reflective posterior hyaloid that may degrade the accuracy of Inner-Segment/Outer-Segment(IS/OS) junction layer and Bruch's membrane (BM) detection; 2) coarse segmentation of BM and IS/OS, by the use of 3-D curvelet transform and graph theory; 3) fine segmentation, in which morphological operators are used to exclude falsely detected elongated structures and get the refined dome-shaped spaces which are constructed between the RPE and Bruch’s membrane.

The proposed method was evaluated on 25 publically available volumetric scans with intermediate AMD from the Age-Related Eye Disease Study 2 (AREDS2) Ancillary SD-OCT Study acquired by using Bioptigen spectral-domain ophthalmic imaging system diagnosed with serous PED. The average true positive volume fraction (TPVF), false positive volume fraction (FPVF), dice coefficient (DC) are 92.13%±4.13%, 0.27%+0.12, 90.12%+2.21 respectively.

An automated method is proposed for 3D serous PED segmentation in SD-OCT images. As an alternative efficient replacement of manual segmentation, the proposed algorithm can reduce time and labor costs and can provide clinicians with proper quantitative information. Using this software, PED size and volume measurements may be a useful tool for quantifying AMD severity in clinical practice.