Automated framework for intraretinal cystoid macular edema segmentation in three-dimensional optical coherence tomography images with macular hole

Abstract

Cystoid macular edema (CME) and macular hole (MH) are the leading causes for visual loss in retinal diseases. The volume of the CMEs can be an accurate predictor for visual prognosis. This paper presents an automatic method to segment the CMEs from the abnormal retina with coexistence of MH in three-dimensional-optical coherence tomography images. The proposed framework consists of preprocessing and CMEs segmentation. The preprocessing part includes denoising, intraretinal layers segmentation and flattening, and MH and vessel silhouettes exclusion. In the CMEs segmentation, a three-step strategy is applied. First, an AdaBoost classifier trained with 57 features is employed to generate the initialization results. Second, an automated shape-constrained graph cut algorithm is applied to obtain the refined results. Finally, cyst area information is used to remove false positives (FPs). The method was evaluated on 19 eyes with coexistence of CMEs and MH from 18 subjects. The true positive volume fraction, FP volume fraction, dice similarity coefficient, and accuracy rate for CMEs segmentation were 81.0%±7.8%, 0.80%±0.63%, 80.9%±5.7%, and 99.7%±0.1%, respectively.

An unsupervised hierarchical approach for automatic intra‐retinal cyst segmentation in spectral‐domain optical coherence tomography images

“…Traditional supervised techniques such as [7][8][9][10][11][12][13] used hand-crafted features and simple classifiers for the segmentation of IRCs. Well-known classifiers like k-nearest neighbor, 7 support vector machine, 8 random forest, 9 kernel regression, 10 graph search-graph cut 11 and AdaBoost, 12 along with a variety of shapes, intensity, spatial and texture-based features have been widely used in these works. Some of these methods like 7,13 need accurate layer segmentation for computing their features, while segmentation of middle retinal layers in the presence of IRCs is a challenging problem.…”

“…In these methods, there are two generations of approaches: traditional and deep learning‐based supervised methods. Traditional supervised techniques such as 7–13 used hand‐crafted features and simple classifiers for the segmentation of IRCs. Well‐known classifiers like k‐nearest neighbor, 7 support vector machine, 8 random forest, 9 kernel regression, 10 graph search‐graph cut 11 and AdaBoost, 12 along with a variety of shapes, intensity, spatial and texture‐based features have been widely used in these works.…”

A Review of Machine Learning Algorithms for Retinal Cyst Segmentation on Optical Coherence Tomography

“…Specifically, their approach involved Gavin’s algorithm in detecting retinal layer boundaries [ 32 ], followed by using probability normalization to refine the initial results. For intra-retinal CME segmentation, a SNR balancing [ 56 ] was performed on each OCT scan to obtain a uniform intensity profile. The method isolated macular hole (MH) and vessel silhouettes before extracting volumetric features of the CME.…”

Future clinical applicability of optical coherence tomography angiography

“…Automated segmentation methods can be based either on fixed model or machine/deep learning. Fixed model‐based methods include peak searching in A‐scan, adaptive thresholding, Markov boundary model, active contour, texture and shape analysis, 3‐D theoretical representation fitting, sparse high order potentials, graph theory and dynamic programming . These techniques are fed with presumed layer structure and intensity information and are hence sensitive to system parameters as well as data formats.…”