OCT影像人工智能读片与医生读片对识别年龄相关性黄斑变性的一致性分析

Consistency analysis of OCT image by artificial intelligence recognition and ophthalmologist's recognition for age-related macular degeneration

摘要:

目的：探讨在临床进行年龄相关性黄斑变性(ARMD)患者眼底光学相干断层扫描(OCT)图像人工智能(AI)读片的可行性。

方法：收集2019-11/2021-11 上海市静安区市北医院眼科门诊患者1 579眼OCT图像共1 579张，并收集眼科医生及AI的读片结果。通过Kappa值进行无ARMD和有ARMD分类结果的一致性分析。

结果：两名眼科医生之间在无ARMD和有ARMD读片结果的Kappa值为0.934； AI结果与眼科医生在无ARMD和有ARMD读片结果的Kappa值为0.738。AI对ARMD识别的灵敏度为73.08%，特异度为95.07%，曲线下区域面积(AUC)为0.841。

结论：AI在基于OCT图像的ARMD识别上与眼科医生有较高的一致性，适用于基层医院完成对ARMD的早期筛查和早期转诊工作。

Abstract:

AIM: To investigate the feasibility of artificial intelligence(AI)in reading retinal optical coherence tomography(OCT)images of age-related macular degeneration(ARMD)in clinic.

METHODS: From November 2019 to November 2021, a total of 1 579 OCT images were collected in the outpatient department, and the imaging results of ophthalmologist and AI were collected. The Kappa consistency test of classification results without ARMD and with ARMD were analyzed.

RESULTS: The Kappa coefficients of the judgement of ophthalmologists about ARMD was 0.934. The Kappa coefficients between AI and ophthalmologists was 0.738. The sensitivity, specificity and area under curve(AUC)of AI to ARMD were 73.08%, 95.07% and 0.841 respectively.

CONCLUSION: AI has a high consistency with ophthalmologists in the recognition of ARMD based on OCT images, which is suitable for primary hospitals to complete the early screening and early referral of ARMD.