早产儿视网膜病变眼底检查方法及眼底图像特征研究进展

Research progress on fundus examination methods and fundus image characteristics of retinopathy of prematurity

摘要:

早产儿视网膜病变(ROP)是可预防的儿童期失明主要原因，具有筛查难、诊断难及客观评估难等特点。ROP筛查有多种方式，包括各种接触或非接触成像设备、基于智能手机的眼底照相、基于人工智能的眼底图像分析等。ROP诊断是基于对ROP整个视网膜眼底的可视化和记录，也是后续筛查、治疗评估的基础。眼底筛查对早期识别非常关键，有助于早期发现和及时转诊。通过分析总结ROP眼底图像特征，可能发现其潜在的特征规律，及时有针对性地预防和治疗。人工智能可促进ROP筛查的自动化、可量化和客观诊断。本文综述目前临床上常用的ROP眼底检查方法及眼底图像特征规律，总结人工智能用于ROP自动诊断的最新研究进展。

Abstract:

Retinopathy of prematurity(ROP)is the primary cause of preventable childhood blindness. It is hard to screen, diagnose and objectively evaluate. There are various modalities for ROP screening, including various contact or non-contact imaging devices, smart phone-based fundus photography, and artificial intelligence-based fundus image analysis. The diagnosis of ROP is based on visualization and recording of the entire retinal fundus of ROP, which is also the basis for subsequent screening, treatment assessment. Fundus screening is critical for early recognition and facilitates early detection and prompt referral. Potential features may be found by analyzing and summarizing the characteristics of ROP fundus images. Subsequently, timely and targeted ROP prevention and treatment could be performed. Artificial intelligence promotes automatic, quantifiable and objective diagnosis of ROP. This article reviews commonly used clinical fundus examination methods and fundus image characteristics of ROP and summarizes the latest research progress on the application of artificial intelligence in the automatic diagnosis of ROP.