Annotated retinal optical coherence tomography images (AROI) database for joint retinal layer and fluid segmentation

Abstract

Optical coherence tomography (OCT) images of the retina provide a structural representation and give an insight into the pathological changes present in age-related macular degeneration (AMD). Due to the three-dimensionality and complexity of the images, manual analysis of pathological features is difficult, time-consuming, and prone to subjectivity. Computer analysis of 3D OCT images is necessary to enable automated quantitative measuring of the features, objectively and repeatedly. As supervised and semi-supervised learning-based automatic segmentation depends on the training data and quality of annotations, we have created a new database of annotated retinal OCT images -the AROI database. It consists of 1136 images with annotations for pathological changes (fluid accumulation and related findings) and basic structures (layers) in patients with AMD. Inter-and intra-observer errors have been calculated in order to enable the validation of developed algorithms in relation to human variability. Also, we have performed the automatic segmentation with standard U-net architecture and two state-of-the-art architectures for medical image segmentation to set a baseline for further algorithm development and to get insight into challenges for automatic segmentation. To facilitate and encourage further research in the field, we have made the AROI database openly available.

Association Between Visual Acuity and Residual Retinal Fluid Following Intravitreal Anti–Vascular Endothelial Growth Factor Treatment for Neovascular Age-Related Macular Degeneration

“…At the same time, the presence of SRF has been reported to be associated with milder nAMD and may protect against macular atrophy . Treatment-resistant SRF appears to not be associated with poorer visual acuity either . It has been hypothesized that SRF may be associated with drusen or drusenoid pigment epithelium detachments in certain cases as opposed to neovascularization .…”

“…Using retinal thickness alone to guide treatment decisions may not be sufficient, given that retinal thickness can be influenced by either SRF or IRF. With recent advancements in artificial intelligence segmentation algorithms, it may be possible to automate parts of the clinical decision-making process based on computerized identification of retinal fluid and/or quantification of fluid volumes …”

Simultaneous Segmentation of Layers and Fluids in Retinal OCT Images

“…K-fold cross validation was used for training, validation and testing. For fair comparison, we used the same data splits as in the baseline model [13]: Each fold consists of B-scans from 4 patients. For examples, the first fold consists of patient 1,2,3 and 4, and then patient 5,6,7 and 8 for the second fold and so on.…”

Retinal layer and fluid segmentation in optical coherence tomography images using a hierarchical framework

“…13 Although most of the methods developed for retinal layer and fluid segmentation are trained and evaluated on normal OCT exams and on OCT exams with signs of DME, there are some methods designed to handle abrupt changes in attenuation coefficients within a layer and to analyze topology-disrupting anomalies that are typical of other pathologies (like central serous retinopathy and age-related macular degeneration). 14,15 Fig. 1 Example of an OCT B-scan showing signs of DME.…”