Automated assessment of the smoothness of retinal layers in optical coherence tomography images using a machine learning algorithm

Abstract

Quantifying the smoothness of different layers of the retina can potentially be an important and practical biomarker in various pathologic conditions like diabetic retinopathy. The purpose of this study is to develop an automated machine learning algorithm which uses support vector regression method with wavelet kernel and automatically segments two hyperreflective retinal layers (inner plexiform layer (IPL) and outer plexiform layer (OPL)) in 50 optical coherence tomography (OCT) slabs and calculates the smoothness index (SI). The Bland–Altman plots, mean absolute error, root mean square error and signed error calculations revealed a modest discrepancy between the manual approach, used as the ground truth, and the corresponding automated segmentation of IPL/ OPL, as well as SI measurements in OCT slabs. It was concluded that the constructed algorithm may be employed as a reliable, rapid and convenient approach for segmenting IPL/OPL and calculating SI in the appropriate layers.

Assessment of area and structural irregularity of retinal layers in diabetic retinopathy using machine learning and image processing techniques

“…Emerging optical coherence tomography (OCT) technologies, namely spectral domain OCT (SD-OCT) and swept-source OCT (SS-OCT), offer improved visibility of retinal layers, hence facilitating a more precise assessment of irregularity and area of retinal layers 28 , 29 . Nevertheless, it is imperative to tackle obstacles such as segmentation errors and discrepancies across optical coherence tomography (OCT) instruments in order to provide uniform measurements across diverse clinical environments 17 , 19 , 21 .…”

“…The retinal layers were segmented automatically using the method described in our recently published study 17 . The automated procedure entailed mitigating artifacts and enhancing image quality through the utilization of nonlocal algorithms, notably Gaussian filters with a kernel size of 11 × 11.…”

“…Subsequently, Gabor filters with two distinct kernel sizes of 6 × 6 were employed to specifically focus on the retinal area. The details were previously described in our published paper, where we supplied the Gabor filter formula and associated parameters in a companion table 17 . In all horizontal OCT scans, an unwanted region known as the optic disc is evident.…”

“…Similarly, the presence and localization of hyperreflective foci (HF) within the retina may have a similar prognostic value 16 . The utilization of new OCT biomarkers could enhance risk stratification, aid in characterizing disease morphology, optimize the use of specific therapies such as anti-vascular endothelial growth factor (VEGF) therapy, improve prognosis counseling, and provide more precise inclusion criteria for therapeutic trials 17 , 18 .…”

“…There are various methods to accomplish this task, including manual, semi-automated, or fully automated approaches 19 – 23 . In a recent study, an automated technique was developed to serve as a dependable, expeditious, and user-friendly method for segmenting the inner plexiform layer (IPL) and outer plexiform layer (OPL), as well as estimating the smoothness index (SI) inside these specific layers 17 . Automated segmentation methods are advantageous due to their ability to operate continuously and independently of user experience, thereby mitigating the time-consuming and low repeatability nature of manual segmentation procedures 17 , 19 , 23 .…”