Deep learning algorithms to segment and quantify the choroidal thickness and vasculature in swept-source optical coherence tomography images

Abstract

Accurate segmentation of choroidal thickness (CT) and vasculature is important to better analyze and understand the choroid-related ocular diseases. In this paper, we proposed and implemented a novel and practical method based on the deep learning algorithms, residual U-Net, to segment and quantify the CT and vasculature automatically. With limited training data and validation data, the residual U-Net was capable of identifying the choroidal boundaries as precise as the manual segmentation compared with an experienced operator. Then, the trained deep learning algorithms was applied to 217 images and six choroidal relevant parameters were extracted, we found high intraclass correlation coefficients (ICC) of more than 0.964 between manual and automatic segmentation methods. The automatic method also achieved great reproducibility with ICC greater than 0.913, indicating good consistency of the automatic segmentation method. Our results suggested the deep learning algorithms can accurately and efficiently segment choroid boundaries, which will be helpful to quantify the CT and vasculature.

Impact of Penetration and Image Analysis in Optical Coherence Tomography on the Measurement of Choroidal Vascularity Parameters

“…Instead of manually outlining the choroidal layers, we used custom-developed Python-based software and developed a ResNet-UNet to automatically identify the layers and measured the LV, TCV, and CVI within 6 mm of the center fovea. Our automatic method was validated using a manual method in our previous study, 15 and is more efficient and versatile for routine clinical practice and research. The study further reports the impact of penetration, analysis image method, and scanning speed in the acquisition of choroidal vascularity parameters in healthy and young cohort, suggesting the consideration of these factors in the research and clinical work.…”

“…10,11 Previous studies have analyzed CVI using different image processing methods and various OCT instruments. 10,[12][13][14][15] It can be assumed that more OCT instruments with higher penetration and more automatic segmentation will be available for future research and routine clinical use. One may assume that higher penetration with less manual intervention on the image analysis may lead to better information for choroid vascularity measurements.…”

Acute hyperglycemia compromises the responses of choroidal vessels using swept-source optical coherence tomography during dark and light adaptations

“…Since Agrawal et al. first proposed the Niblack image binarization technique for EDI-OCT images, the choroid can be differentiated into luminal and stromal components ( 10 , 14 ). Through the acquisition of more detailed information on the choroidal structures using the deep-learning-based automatic segmentation and modified image binarization technique, our study demonstrated the effect of acute hyperglycemia on the choroidal components and CVI while modulating light adaptation in healthy participants.…”

“…A custom algorithm based on image binarization and AI segmentation of SS-OCT ( 14 ) was used to obtain the choroidal structural parameters and vascularity index, which has been described in detail in our previous studies ( 9 , 21 ). Briefly, the upper and lower boundaries of the choroid were automatically detected using an algorithm based on the deep learning network implemented in MATLAB 2017a (Mathworks, Inc., Natick, MA, USA).…”

“…This method has been validated with good reproducibility and also applied in research in many diseases, such as DR ( 11 ), central serous chorioretinopathy (CSC) ( 12 ), age-related macular degeneration (AMD) ( 13 ). Recently, our group developed an AI segmentation combined with Niblack’s automatic local threshold to obtain choroid volume parameters ( 14 ). Therefore, choroidal structural parameters, including luminal volume (LV), stromal volume (SV), total choroidal volume (TCV), and CVI, defined as the ratio of LV to TCV, may be potential quantitative indicators for characterizing the choroidal response while modulating light adaptation and blood glucose.…”