Abstract
We developed a fully automated system using a convolutional neural network (CNN) for total retina segmentation in optical coherence tomography (OCT) that is robust to the presence of severe retinal pathology. A generalized U-net network architecture was introduced to include the large context needed to account for large retinal changes. The proposed algorithm outperformed qualitative and quantitatively two available algorithms. The algorithm accurately estimated macular thickness with an error of 14.0 ± 22.1 µm, substantially lower than the error obtained using the other algorithms (42.9 ± 116.0 µm and 27.1 ± 69.3 µm, respectively). These results highlighted the proposed algorithm's capability of modeling the wide variability in retinal appearance and obtained a robust and reliable retina segmentation even in severe pathological cases.
Highlights
Optical coherence tomography (OCT) has become a key diagnostic imaging technique for the diagnosis of retinal diseases [1]
Retinal thickness or central macular thickness (CMT) as measured in OCT has been demonstrated to correlate with pathological changes and treatment outcome for a variety of ocular diseases [5,6,7]
We propose an a machine learning algorithm for total retina thickness segmentation using a convolutional neural network (CNN) that is robust to severe disruptive retinal pathology
Summary
Optical coherence tomography (OCT) has become a key diagnostic imaging technique for the diagnosis of retinal diseases [1]. Every pixel in each B-scan is directly classified as belonging to the retina or not, avoiding assumptions about the retinal structure and its layered organization This absence of constraints and a priori knowledge allows for a flexible model that is capable of capturing the wide range of possible variations in retinal appearance when disruptive pathology is present. The proposed CNN architecture builds upon U-net [47], a fully convolutional network for spatially dense prediction tasks This architecture implements a multi-scale analysis of an image, allowing spatially distant information to be included in the classification decision but maintaining pixel accurate localization avoiding the need of an additional step based on graph search. As initial segmentations were not provided to this observer, only A-scans intersecting with the 1 mm circle surrounding the fovea were annotated in order to reduce the annotation workload
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.