Automatic identification and characterization of the epiretinal membrane in OCT images

Abstract

Optical coherence tomography (OCT) is a medical image modality that is used to capture, non-invasively, high-resolution cross-sectional images of the retinal tissue. These images constitute a suitable scenario for the diagnosis of relevant eye diseases like the vitreomacular traction or the diabetic retinopathy. The identification of the epiretinal membrane (ERM) is a relevant issue as its presence constitutes a symptom of diseases like the macular edema, deteriorating the vision quality of the patients. This work presents an automatic methodology for the identification of the ERM presence in OCT scans. Initially, a complete and heterogeneous set of features was defined to capture the properties of the ERM in the OCT scans. Selected features went through a feature selection process to further improve the method efficiency. Additionally, representative classifiers were trained and tested to measure the suitability of the proposed approach. The method was tested with a dataset of 285 OCT scans labeled by a specialist. In particular, 3,600 samples were equally extracted from the dataset, representing zones with and without ERM presence. Different experiments were conducted to reach the most suitable approach. Finally, selected classifiers were trained and compared using different metrics, providing in the best configuration an accuracy of 89.35%.

Automatic Identification and Representation of the Cornea–Contact Lens Relationship Using AS-OCT Images

“…In line with that, and in order to take a deeper look into a particularly similar domain to ours, it is noticeable that, recently, snakes were satisfactorily used for the segmentation of retinal layers in OCT images [22,23,24]. In particular, González-López et al [24] proposed a strategy based on snakes for the segmentation of the main retinal layers where good results were obtained from the designed experiments, outperforming previous existing proposals in this specific domain, including a geometric level set formulation of an active contour model [25] and a deep learning approach [26], among others.…”

“…In particular, snakes were used for the segmentation of boundaries in different medical imaging modalities demonstrating the suitability of this strategy in terms of simplicity of implementation, robustness and efficiency of required computational resources. In particular, as representative examples, similar approaches were satisfactorily applied in OCT, a very related domain about the analysis of the eye fundus, for the segmentation of the retinal layers [22,24,37] or the measurement of the retinal thickness [23], among other applications. A snake consists of a deformable model under the influence of internal forces that restrict the shape of the model (for example, its smoothness) and external forces that push the curve towards characteristics that are present in the image (for example, the edges).…”

Automatic Identification and Intuitive Map Representation of the Epiretinal Membrane Presence in 3D OCT Volumes

“…To validate this hypothesis, we used a complete and heterogeneous set of 452 intensity-, texture-, and domain-related features with high discriminative value that have been proved to maximize the amount of available information for each point of interest [33], briefly illustrated in Table 1:…”

“…Wilkins et al [31] automatically measured the central macular thickness by using manually defined boundaries around the studied region in individual 2D OCT slices. Similarly, other previous works on the matter [32,33] proposed an automatic methodology to identify the ERM’s presence in individual 2D OCT histological sections by extracting a heterogeneous set of features from the ILM layer region. To date, no proposed methodology entirely exploits the information of the 3D OCT volumes that may be obtained from the macular region of an individual patient.…”

Automatic Segmentation and Intuitive Visualisation of the Epiretinal Membrane in 3D OCT Images Using Deep Convolutional Approaches

“…As we can observe from the literature, the presented methods only considered the problem of automatic identification, a problem that is easier to address than precise segmentation. In this regard, only Baamonde et al [34] presented a fully automatic method for ERM segmentation by performing a sliding window classification over the ILM layer, determining the presence or not of the ERM disease in 3D OCT volumes. In particular, the authors added a post-processing stage over the classification results to integrate information from the surrounding layers into the segmentation maps.…”