Abstract
Purpose :
In age-related macular degeneration (AMD) the presence of intraretinal cysts is an important prognostic biomarker. Optical coherence tomography imaging has shown to be capable of accurately visualizing and characterizing the three-dimensional shape and extent of cysts. The detection and segmentation of intraretinal cysts is highly beneficial for the prediction of treatment outcome and the assessment of the treatment progression. To aid the clinician with quantified information regarding cysts, we developed a fully automated system capable of detecting and delineating intraretinal cysts in 3D optical coherence tomography images.
Methods :
A total of 30 OCT volumes acquired with four different OCT scanners were provided by the OPTIMA cyst segmentation challenge containing a wide variety of cysts. A pixel classifier based on a multiscale convolutional neural network (CNN) was developed to predict if an image voxel belongs to a cyst by considering a small neighborhood around the voxel of interest. The CNN comprises of 3 parallel subnetworks to include information at different image scales. Each subnetwork consists of 4 convolutional layers, a max pooling layers and fully connected layer. The output of the subnetworks are merged in a final fully connected layer. After providing the system with enough training samples, the network can automatically detect and segment cysts in OCT volumes. The obtained segmentations were compared to manual delineations made by two experienced human graders for the central 3 mm surrounding the fovea.
Results :
The spatial overlap agreement on the obtained volumes, measured by the Dice similarity coefficient, compared to manual delineations for images acquired with a Cirrus, Nidek, Spectralis or Topcon scanner, were 0.62, 0.57, 0.60 and 0.71, respectively. Sensitivity/positive predictive value pairs for the before mentioned OCT vendors are 0.57/0.69, 0.99/0.35, 0.77/0.54 and 0.77/0.69.
Conclusions :
An image analysis algorithm for the automatic segmentation of intraretinal cysts in OCT images was developed. The proposed algorithm is able to detect and quantify the three dimensional shape and extent of cysts in a fast and reproducible manner, allowing accurate assessment of volume changes and treatment outcome.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.