From the Hamilton Glaucoma Center, Shiley Eye Institute, and The Viterbi Family Department of Ophthalmology, University of California, San Diego, La Jolla, California, USA.
From the Hamilton Glaucoma Center, Shiley Eye Institute, and The Viterbi Family Department of Ophthalmology, University of California, San Diego, La Jolla, California, USA.
Am J Ophthalmol. 2022 Apr;236:298-308. doi: 10.1016/j.ajo.2021.11.008. Epub 2021 Nov 13.
To compare convolutional neural network (CNN) analysis of en face vessel density images to gradient boosting classifier (GBC) analysis of instrument-provided, feature-based optical coherence tomography angiography (OCTA) vessel density measurements and OCT retinal nerve fiber layer (RNFL) thickness measurements for classifying healthy and glaucomatous eyes.
Comparison of diagnostic approaches.
A total of 130 eyes of 80 healthy individuals and 275 eyes of 185 glaucoma patients with optic nerve head (ONH) OCTA and OCT imaging were included. Classification performance of a VGG16 CNN trained and tested on entire en face 4.5 × 4.5-mm radial peripapillary capillary OCTA ONH images was compared to the performance of separate GBC models trained and tested on standard OCTA and OCT measurements. Five-fold cross-validation was used to test predictions for CNNs and GBCs. Areas under the precision recall curves (AUPRC) were calculated to control for training/test set size imbalance and were compared.
Adjusted AUPRCs for GBC models were 0.89 (95% CI = 0.82, 0.92) for whole image vessel density GBC, 0.89 (0.83, 0.92) for whole image capillary density GBC, 0.91 (0.88, 0.93) for combined whole image vessel and whole image capillary density GBC, and 0.93 (0.91, 095) for RNFL thickness GBC. The adjusted AUPRC using CNN analysis of en face vessel density images was 0.97 (0.95, 0.99) resulting in significantly improved classification compared to GBC OCTA-based results and GBC OCT-based results (P ≤ 0.01 for all comparisons).
Deep learning en face image analysis improves on feature-based GBC models for classifying healthy and glaucoma eyes.
比较基于卷积神经网络（CNN）的血管密度图像分析与基于梯度提升分类器（GBC）的仪器提供的、基于特征的光学相干断层扫描血管造影（OCTA）血管密度测量和 OCT 视网膜神经纤维层（RNFL）厚度测量在分类健康眼和青光眼眼中的应用。
诊断方法比较。
共纳入 80 名健康个体的 130 只眼和 185 名青光眼患者的 275 只眼，这些患者均有视神经头（ONH）OCTA 和 OCT 成像。在整个 4.5×4.5mm 放射状视盘毛细血管 OCTA ONH 图像上训练和测试的 VGG16 CNN 的分类性能与分别在标准 OCTA 和 OCT 测量上训练和测试的单独 GBC 模型的性能进行了比较。使用五折交叉验证来测试 CNN 和 GBC 的预测。计算精度召回曲线下面积（AUPRC）以控制训练/测试集大小的不平衡，并进行比较。
GBC 模型的调整后 AUPRC 分别为：全图像血管密度 GBC 为 0.89（95%置信区间：0.82，0.92），全图像毛细血管密度 GBC 为 0.89（0.83，0.92），全图像血管和全图像毛细血管密度 GBC 联合为 0.91（0.88，0.93），RNFL 厚度 GBC 为 0.93（0.91，0.95）。使用基于 CNN 的血管密度图像分析的调整后 AUPRC 为 0.97（0.95，0.99），与 GBC-OCTA 结果和 GBC-OCT 结果相比，分类性能显著提高（所有比较 P≤0.01）。
深度学习基于视盘的图像分析提高了基于特征的 GBC 模型在分类健康眼和青光眼眼中的性能。
