【摘要】 早产儿视网膜病变(ROP)是导致儿童可预防性失明的主要病因之一，其诊疗受制于医疗资源紧张和专业人才匮乏等困境。近年来，人工智能(AI)技术的发展为ROP智能筛查、诊断和管理带来了新的可能，主要包括对病变分区、分期与附加病变的自动评估，复发风险预测模型的构建，以及推动远程医疗在经济与可及性方面的优化。ROP相关AI模型经历了由传统机器学习向深度学习演进的过程，推动了算法性能的持续提升。尽管AI在ROP中的应用前景广阔，但其临床落地仍面临诸多挑战，包括图像数据的异质性、模型泛化能力不足、缺乏标准化流程以及真实世界验证的有限性。本文旨在评述当前AI技术在ROP领域的关键进展，并指出未来研究需重点突破的瓶颈问题，以实现AI辅助ROP诊疗的高质量发展。更多还原

【Abstract】 Retinopathy of prematurity(ROP) is one of the leading causes of preventable childhood blindness. Its diagnosis and treatment are constrained by challenges such as limited medical resources and a shortage of specialized professionals. In recent years, advancements in artificial intelligence(AI) technology have introduced new possibilities for ROP screening, diagnosis, and management. These include automated assessment of lesion zone, stage, and plus disease; the development of predictive models for recurrence risk; and the promotion of telemedicine to enhance cost-effectiveness and accessibility. From a technical perspective, AI models for ROP have evolved from traditional machine learning to deep learning, driving continuous improvements in algorithm performance. Although the application of AI in ROP is widely regarded as promising, its clinical implementation still faces numerous challenges. These include heterogeneity in image data, insufficient model generalization capabilities, lack of standardized protocols, and limited realworld validation. This article aims to review the key recent advances in AI technology within the ROP field and identify critical bottlenecks requiring focused research breakthroughs. The goal is to facilitate the high-quality development of AI-assisted ROP diagnosis and treatment.更多还原