Deep Learning Estimation of 10-2 and 24-2 Visual Field Metrics Based on Thickness Maps from Macula OCT
Deep Learning Estimation of 10-2 and 24-2 Visual Field Metrics Based on Thickness Maps from Macula OCT
复制标题
DOI：
10.1016/j.ophtha.2021.04.022
复制
发表时间：
2021-10-20
期刊：
影响因子：
13.7
通讯作者：
Zangwill, Linda M.
中科院分区：
文献类型：
--
作者：
Christopher, Mark;Bowd, Christopher;Zangwill, Linda M.
目的：开发从黄斑中心光谱域(SD)OCT图像估计视功能的深度学习系统。设计：一种诊断技术的评估。参与者：从645名健康和青光眼受试者(1222只眼)收集2408对10-2视野(VF)SD OCT和2999对24-2 VF SD OCT对。方法：对深度学习模型进行训练，以估计10-2和24-2 VF平均偏差(MD)和模式标准差(PSD)。用视网膜神经纤维层、神经节细胞层、内网状层、神经节细胞-内网层、神经节细胞复合体[GCC]和视网膜6层的厚度数据训练单独和组合的DL模型。主要观察指标：深度学习模型的R2和MAE与10-2和24-2 VF测量结果进行比较。结果：估计10-2的组合DL模型获得的R-2为0.82(95%可信区间[CI]，0.68~0.89)，PSD和MAES分别为1.9分贝(95%CI，1.6-2.4分贝)和1.5分贝(95%CI)，1.2-1.9分贝)。这明显好于10-2MD(0.61[95%CI，0.47-0.71]和3.0dB[95%CI，2.5-3.5dB])和10-2PSD(0.46[95%CI，0.31-0.60]和2.3dB[95%CI，1.8-2.7dB])的平均厚度估计。估计24-2的联合DL模型获得MD和PSD的R-2分别为0.79(95%CI，0.72-0.84)和0.68(95%CI，0.53-0.79)，MAES分别为2.1dB(95%CI，1.8-2.5d B)和1.5dB(95%CI，1.3-1.9d B)。这明显好于24-2 MD(0.41[95%CI，0.26-0.57]和3.4dB[95%CI，2.7-4.5dB])和24-2 PSD(0.38[95%CI，0.20-0.57]和2.4dB[95%CI，2.0-2.8dB])的平均厚度估计。GCIPL(R2=0.79)和GCC(R2=0.75)分别具有最高的估计10-2和24-2 MD的性能。结论：深度学习模型改善了SD OCT成像对功能损失的估计。准确的评估可以帮助临床医生对患者进行个性化的VF测试。(C)2021年由美国眼科学会颁发。
Purpose: To develop deep learning (DL) systems estimating visual function from macula-centered spectraldomain (SD) OCT images.Design: Evaluation of a diagnostic technology.Participants: A total of 2408 10-2 visual field (VF) SD OCT pairs and 2999 24-2 VF SD OCT pairs collected from 645 healthy and glaucoma subjects (1222 eyes).Methods: Deep learning models were trained on thickness maps from Spectralis macula SD OCT to estimate 10-2 and 24-2 VF mean deviation (MD) and pattern standard deviation (PSD). Individual and combined DL models were trained using thickness data from 6 layers (retinal nerve fiber layer [RNFL], ganglion cell layer [GCL], inner plexiform layer [IPL], ganglion cell-IPL [GCIPL], ganglion cell complex [GCC] and retina). Linear regression of mean layer thicknesses were used for comparison.Main Outcome Measures: Deep learning models were evaluated using R2 and mean absolute error (MAE) compared with 10-2 and 24-2 VF measurements.Results: Combined DL models estimating 10-2 achieved R-2 of 0.82 (95% confidence interval [CI], 0.68-0.89) for MD and 0.69 (95% CI, 0.55-0.81) for PSD and MAEs of 1.9 dB (95% CI, 1.6-2.4 dB) for MD and 1.5 dB (95% CI, 1.2-1.9 dB) for PSD. This was significantly better than mean thickness estimates for 10- 2 MD (0.61 [95% CI, 0.47-0.71] and 3.0 dB [95% CI, 2.5-3.5 dB]) and 10-2 PSD (0.46 [95% CI, 0.31-0.60] and 2.3 dB [95% CI, 1.8-2.7 dB]). Combined DL models estimating 24-2 achieved R-2 of 0.79 (95% CI, 0.72-0.84) for MD and 0.68 (95% CI, 0.53-0.79) for PSD and MAEs of 2.1 dB (95% CI, 1.8-2.5 dB) for MD and 1.5 dB (95% CI, 1.3-1.9 dB) for PSD. This was significantly better than mean thickness estimates for 24-2 MD (0.41 [95% CI, 0.26-0.57] and 3.4 dB [95% CI, 2.7-4.5 dB]) and 24-2 PSD (0.38 [95% CI, 0.20-0.57] and 2.4 dB [95% CI, 2.0-2.8 dB]). The GCIPL (R-2 = 0.79) and GCC (R-2 = 0.75) had the highest performance estimating 10-2 and 24-2 MD, respectively.Conclusions: Deep learning models improved estimates of functional loss from SD OCT imaging. Accurate estimates can help clinicians to individualize VF testing to patients. (C) 2021 by the American Academy of Ophthalmology.