文献速递眼底病与葡萄膜炎

模块化多任务深度学习框架预测新生血管性年龄相关性黄斑变性治疗启动时机

Modular multi-task deep learning framework for prediction of treatment initiation in neovascular age-related macular degeneration modular AI for NVAMD treatment initiation study-MANTIS.
PubMed 文献 · Br J Ophthalmol · 2026-09-01

该回顾性研究开发了基于OCT图像和视力评分的全自动深度学习系统,用于确定抗VEGF治疗的最佳启动时间,并估计外层视网膜厚度作为影像生物标志物。研究纳入122例患者(207眼),分为未进展为新生血管性AMD和进展并接受治疗两个队列。任务一模型与临床决策的一致性评估显示AUC为0.73,准确率0.87,敏感度0.69,特异度0.92;任务二估计外层视网膜厚度的NMAE为0.12。该框架为确定最佳治疗时机提供了全自动方法,支持早期干预以保留视力。

要点

  • 深度学习模型可基于OCT和视力预测新生血管性AMD抗VEGF治疗启动,AUC 0.73。
  • 模型估计外层视网膜厚度的归一化平均绝对误差为0.12。
  • 该框架支持早期干预以保留新生血管性AMD患者视力。

核心结论

  • 该深度学习框架为确定新生血管性AMD最佳治疗启动时机提供了全自动方法,支持早期干预以保留视力。

研究设计与样本

  • 回顾性研究,纳入两个队列:未进展为新生血管性AMD的中期干性AMD患者和进展并接受抗VEGF治疗的患者,受试者至少有3次连续随访。
  • 共纳入122例患者(207眼),数据集1有49例(69眼,平均年龄84岁,51%女性),数据集2有100例(138眼,平均年龄82岁,54%女性)。
  • 定义两项任务:任务一开发深度学习模型识别抗VEGF治疗需求;任务二估计从Bruch膜到外 plexiform 层的外层视网膜厚度作为影像生物标志物。
  • 使用卷积神经网络提取与外层视网膜厚度相关的OCT特征,并在 separate 数据集上训练两个长短期记忆模型分别处理任务一和任务二。

主要发现

  • 任务一模型与临床抗VEGF启动决策的一致性评估:AUC 0.73(95%CI 0.59-0.87),AUPRC 0.51(95%CI 0.35-0.66),准确率0.87(95%CI 0.76-0.97),精确率0.73(95%CI 0.59-0.87),敏感度0.69(95%CI 0.54-0.83),特异度0.92(95%CI 0.84-1.00)。
  • 任务二估计外层视网膜厚度的归一化平均绝对误差(NMAE)为0.12(95%CI 0.04-0.20)。

临床怎么用

  • 该全自动深度学习框架可辅助临床医生确定新生血管性AMD患者启动抗VEGF治疗的最佳时机。
  • 模型基于OCT图像和视力评分,可支持早期干预决策以保留患者视力。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

PURPOSE: The purpose of this study is to develop a fully automated system for determining the optimal time to begin anti-vascular endothelial growth factor (anti-VEGF) treatment, based on retinal changes observed in optical coherence tomography (OCT) images and visual acuity (VA) scores. METHODS: This retrospective study included two cohorts: patients with intermediate dry age-related macular degeneration (AMD) who did not progress to neovascular AMD (NVAMD) and those who progressed and received anti-VEGF treatment. Subjects had ≥3 consecutive visits. Two tasks were defined: (1) develop a deep learning (DL) model to identify the need for anti-VEGF treatment and (2) estimate outer retina thickness (ORT) from Bruch's membrane to the outer plexiform layer as an imaging biomarker. A convolutional neural network extracted OCT features related to ORT, and two long short-term memory models were trained on separate datasets to address Task I and II treatments. RESULTS: A total of 122 patients (207 eyes) were included: 49 in dataset 1 (69 eyes; mean (SD) age 84 (8); 25 (51%) female) and 100 in dataset 2 (138 eyes; 82 (7); 54 (54%) female). Task I performance was evaluated by the model's alignment with clinical decisions on anti-VEGF initiation, and demonstrated area under the receiver operating curve of 0.73 (95% CI 0.59 to 0.87), area under the precision-recall curve of 0.51 (95% CI 0.35 to 0.66), an accuracy of 0.87 (95% CI 0.76 to 0.97), a precision of 0.73 (95% CI 0.59 to 0.87), a sensitivity of 0.69 (95% CI 0.54 to 0.83) and a specificity of 0.92 (95% CI 0.84 to 1.00). Task II achieved NMAE 0.12 (95% CI 0.04 to 0.20). CONCLUSIONS: This DL framework provides a fully automated approach for determining optimal treatment timing, supporting earlier intervention to preserve VA in NVAMD patients. TRANSITIONAL RELEVANCE: An accurate DL framework that can help with identifying the best time to initiate therapy for NVAMD.

来源:Br J Ophthalmol,查看原文。版权归原作者所有。

引用

  1. Gao Q, Kuo D, Amason J, et al. Modular multi-task deep learning framework for prediction of treatment initiation in neovascular age-related macular degeneration modular AI for NVAMD treatment initiation study-MANTIS. Br J Ophthalmol. 2026. PMID: 42680556. doi:10.1136/bjo-2025-328327

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。