Semi-Automated Labelling of Cystoid Macular Edema in OCT Scans

The analysis of retinal Spectral-Domain Optical Coherence Tomography (SD-OCT) images by trained medical professionals can be used to provide useful insights into various diseases. It is the most popular method imaging due its non-invasive nature and information it provides for making an accurate diagnosis. A deep learning approach automating segmentation Cystoid Macular Edema (fluid) in OCT B-Scan images was developed that consequently volumetric analysis scans. This solution a fast semantic network that makes use shortened encoder-decoder UNet like architecture with integrated Dense ASPP module Attention Gate producing refined retinal fluid map. Our system evaluated against both publicly privately available datasets; on former network achieved Dice coefficient 0.804, thus making current best performing this dataset, very small and challenging private score 0.691. Due lack publicly available data in domain, Graphical User Interface aims semi-automate the labelling process also created, thus greatly simplifying of the dataset creation potentially leading increase labelled production.