Deep learning-based estimation of axial length and subfoveal choroidal thickness from color fundus photographs

• Verfasst von: Dong, Li [VerfasserIn]
• Verfasst von: Hu, Xin Yue [VerfasserIn]
• Verfasst von: Yan, Yan Ni [VerfasserIn]
• Verfasst von: Zhang, Qi [VerfasserIn]
• Verfasst von: Zhou, Nan [VerfasserIn]
• Verfasst von: Shao, Lei [VerfasserIn]
• Verfasst von: Wang, Ya Xing [VerfasserIn]
• Verfasst von: Xu, Jie [VerfasserIn]
• Verfasst von: Lan, Yin Jun [VerfasserIn]
• Verfasst von: Li, Yang [VerfasserIn]
• Verfasst von: Xiong, Jian Hao [VerfasserIn]
• Verfasst von: Liu, Cong Xin [VerfasserIn]
• Verfasst von: Ge, Zong Yuan [VerfasserIn]
• Verfasst von: Jonas, Jost B. [VerfasserIn]
• Verfasst von: Wei, Wen Bin [VerfasserIn]
• Titel: Deep learning-based estimation of axial length and subfoveal choroidal thickness from color fundus photographs
• Verf.angabe: Li Dong, Xin Yue Hu, Yan Ni Yan, Qi Zhang, Nan Zhou, Lei Shao, Ya Xing Wang, Jie Xu, Yin Jun Lan, Yang Li, Jian Hao Xiong, Cong Xin Liu, Zong Yuan Ge, Jost B. Jonas and Wen Bin Wei
• E-Jahr: 2021
• Jahr: 09 April 2021
• Umfang: 8 S.
• Fussnoten: Gesehen am 12.07.2021
• Titel Quelle: Enthalten in: Frontiers in cell and developmental biology
• Ort Quelle: Lausanne : Frontiers Media, 2013
• Jahr Quelle: 2021
• Band/Heft Quelle: 9(2021), Artikel-ID 653692, Seite 1-8
• ISSN Quelle: 2296-634X
• DOI: doi:10.3389/fcell.2021.653692
• Datenträger: Online-Ressource
• Sprache: eng
• Sach-SW: Axial length
• Sach-SW: Convolution Neural Network
• Sach-SW: deep learning
• Sach-SW: fundus image
• Sach-SW: fundus photography
• Sach-SW: Subfoveal choroidal thickness
• K10plus-PPN: 1762709384

Abstract

This study aimed to develop an automated computer-based algorithm to predict axial length and subfoveal choroidal thickness (SFCT) based on color fundus photographs. In the population-based Beijing Eye Study 2011, we took fundus photographs and measured SFCT by optical coherence tomography (OCT) and axial length by optical low-coherence reflectometry. Using 6,394 color fundus images taken from 3,468 participants, we trained and evaluated a deep learning-based algorithm for estimation of axial length and SFCT. The algorithm had a mean absolute error (MAE) for predicting axial length and SFCT of 0.56 mm (95% confidence interval (CI):0.53,0.61) and 49.20 μm (95% CI:45.83,52.54), respectively. Predicted values and measured data showed coefficients of determination of r2=0.59 (95% CI:0.50,0.65) for axial length and r2=0.62 (95% CI: 0.57,0.67) for SFCT. Bland-Altman plots revealed a mean difference in axial length and SFCT of -0.16mm (95% CI:-1.60,1.27mm) and of -4.40 μm (95%CI:-131.8,122.9 μm), respectively. For the prediction of axial length, heat map analysis showed that signals predominantly from overall of the macular region, the foveal region and the extrafoveal region were used in eyes with an axial length of 26 mm, respectively. For the prediction of SFCT, the CNN used mostly the central part of the macular region, the fovea or perifovea, independently of the SFCT. Our study shows deep learning-based algorithms may be helpful in estimating axial length and SFCT based on conventional color fundus images. They may be a further step in the semi-automatic assessment of the eye.