Segmentation of Intra-Retinal Cysts from Optical Coherence Tomography Images using a Fully Convolutional Neural Network Model

Abstract

Optical Coherence Tomography (OCT) is an imaging modality that is used extensively for ophthalmic diagnosis, near-histological visualization and quantification of retinal abnormalities such as cysts, exudates, retinal layer disorganization, etc. Intra-retinal cysts (IRCs) occur in several macular disorders such as, diabetic macular edema, retinal vascular disorders, age-related macular degeneration, and inflammatory disorders. Automated segmentation of IRCs poses challenges owing to variations in the acquisition system scan intensities, speckle noise, and imaging artifacts. Several segmentation methods have been proposed in the literature for IRC segmentation on vendorspecific OCT images that lack generalizability across imaging systems. In this work, we propose a fully convolutional network (FCN) model for vendor-independent IRC segmentation. The proposed method counteracts image noise variabilities and trains FCN models on OCT sub-images from the OPTIMA cyst segmentation challenge dataset (with four different vendor-specific images, namely, Cirrus, Nidek, Spectralis, and Topcon). Further, optimal data augmentation and model hyper-parametrization is shown to prevent over-fitting for IRC area segmentation. The proposed method is evaluated on the test data set with a recall/precision rate of 0.66/0.79 across imaging vendors. The Dice correlation coefficient of the proposed method outperforms that of the published algorithms in the OPTIMA cyst segmentation challenge with a Dice rate of 0.71 across the vendors.

Existing System

k-Nearest Neighbour (kNN) classifier

Proposed System

This paper makes two key contributions. First, a customizable state-of-the-art FCN model is presented that is capable of automated IRC area segmentation from OCT images across vendor-specific imaging systems. We analyze the sensitivity of model parameters, such as number of layers and kernel dimensions, towards the IRC area segmentation goals. We observe that an optimally parametrized model can achieve higher recall rate of 0.66 while preserving the precision rate of 0.79 across multiple vendor data, when compared to stateof-the-art methods. Second, the importance of OCT image pre-processing by image noise suppression using Gamma noise models, sub-retinal region of interest (ROI) segmentation and optimal data augmentation methods are presented. We observe that image pre-processing and domain-specific data augmentation methods significantly prevent model over-fitting while ensuring generalizability across vendor-specific imaging systems.

Architecture

BLOCK DIAGRAM