Yoshida Masaaki, Kunimatsu-Sanuki Shiho, Omodaka Kazuko, Nakazawa Toru
a Department of Ophthalmology , Tohoku University Graduate School of Medicine , Sendai , Miyagi , Japan.
b Department of Retinal Disease Control , Tohoku University Graduate School of Medicine , Sendai , Miyagi , Japan.
Curr Eye Res. 2018 Jun;43(6):754-761. doi: 10.1080/02713683.2018.1439065. Epub 2018 Feb 16.
This study aimed to calculate a predicted integrated visual field (IVF) based on predicted monocular visual fields (MVFs) derived, with a new method, from wide-scan optical coherence tomography (OCT) data.
Visual field testing used the central (6 × 4) 24 points of the Humphrey Field Analyzer 24-2 program. OCT scans of a corresponding retinal area, centered on the fovea, were divided into a 6 × 4 grid. The thickness of the macular retinal nerve fiber layer (mRNFL), ganglion cell layer + inner plexiform layer (GCIPL), and mRNFL + GCIPL (GCC) was measured in each grid area. Next, a support vector machine was used to create a MVF prediction model, with training data from 101 eyes of 60 glaucoma patients. Then, the prediction model was validated with data from 108 eyes of 54 glaucoma patients, for MVF and IVF. A simulated IVF was created by merging bilateral simulated MVFs.
The overall average of the median 95% prediction interval length for the MVF prediction model (measured in dB) was 10.0, 18.3, and 11.3 for the mRNFL, GCIPL, and GCC, respectively. In the validation data, the overall average root mean squared error (dB) between actual and predicted sensitivity for the IVF was 9.6, 10.5, and 9.5 for the mRNFL, GCIPL, and GCC, respectively, in the 24 grid areas. The intraclass correlation coefficient between average actual and predicted IVF was 0.61, 0.44, and 0.59 in the mRNFL, GCIPL, and GCC, respectively, in the 24 grid areas.
We calculated a predicted IVF based on predicted MVFs that were derived, with a new method, from OCT data and validated the accuracy of the calculated IVF. This technique should improve glaucoma management in cases when standard visual field testing is difficult.
本研究旨在基于通过一种新方法从宽扫描光学相干断层扫描（OCT）数据得出的预测单眼视野（MVF）来计算预测综合视野（IVF）。
视野测试采用Humphrey视野分析仪24 - 2程序的中央（6×4）24个点。以黄斑中心凹为中心的相应视网膜区域的OCT扫描被划分为一个6×4网格。在每个网格区域测量黄斑视网膜神经纤维层（mRNFL）、神经节细胞层 + 内丛状层（GCIPL）以及mRNFL + GCIPL（GCC）的厚度。接下来，使用支持向量机创建一个MVF预测模型，训练数据来自60例青光眼患者的101只眼睛。然后，使用54例青光眼患者的108只眼睛的数据对预测模型进行MVF和IVF验证。通过合并双侧模拟MVF创建模拟IVF。
MVF预测模型的中位数95%预测区间长度（以dB为单位测量）的总体平均值，mRNFL、GCIPL和GCC分别为10.0、18.3和11.3。在验证数据中，24个网格区域内IVF的实际和预测敏感度之间的总体平均均方根误差（dB），mRNFL、GCIPL和GCC分别为9.6、10.5和9.5。24个网格区域内平均实际和预测IVF之间的组内相关系数，mRNFL、GCIPL和GCC分别为0.61、0.44和0.59。
我们基于通过一种新方法从OCT数据得出的预测MVF计算了预测IVF，并验证了计算出的IVF的准确性。在标准视野测试困难的情况下，该技术应能改善青光眼的管理。
