多光谱眼底成像人工智能对视网膜动脉硬化的早期识别效能  

Early Identification Efficiency of Multi-Spectral Fundus Imaging Artificial Intelligence on Retinal Arteriosclerosis

作　　者：马菲妍 张彩霞 冬雪川 Ma FeiYan;Zhang Caixia;Dong Xuechuan(Department of Ophthalmology,Second Hospital of Hebei Medical University,Shijiazhuang Hebei 050005,China;R&D Department,Shenzhen New Industry Ophthalmology New Technology Co.,Ltd.,Shenzhen Guangdong 518055,China)

机构地区：[1]河北医科大学第二医院眼科,河北石家庄050005 [2]深圳市新产业眼科新技术有限公司研发部,广东深圳518055

出　　处：《医疗装备》2022年第3期1-4,共4页Medical Equipment

基　　金：深圳市科技研发资金—深科技创新〔2019〕33号(JSGG20180507182010237);河北省2020年度医学科学研究课题计划(20200069)。

摘　　要：目的探讨基于多光谱眼底成像开发的人工智能(AI)在视网膜动脉硬化早期诊断中的应用价值。方法采用诊断试验的研究方法,以150张经过专家标定的可能患有不同程度视网膜动脉硬化的多光谱眼底图像作为阅片标注的参考标准,比较AI组、高年资眼科医师组、低年资眼科医师组及心血管内科医师组的诊断一致性和阅片速度。结果对于各组的阅片一致性,AI组与高年资眼科医师组、低年资眼科医师组、心血管内科医师组比较的Kendall协调系数分别为0.887、0.853、0.848,P<0.01,诊断结果基本一致;AI组、高年资眼科医师组、低年资眼科医师组、心血管内科医师组的平均单张阅片时间分别为(1.52±0.29)、(14.70±1.74)、(22.36±2.43)、(43.10±7.08)s,4组平均单张阅片时间比较,差异有统计学意义(F=3134.857,P<0.01)。结论通过AI和多光谱眼底成像技术的结合能够提升视网膜动脉硬化的筛查效能,降低开发难度,利于不同资质的医师,尤其是全科医师和年轻医师快速掌握本病的诊断和筛查方法。Objective The application value of developed artificial intelligence(AI)based on multi-spectral fundus imaging in the early diagnosis of retinal arteriosclerosis was explored.Methods Using the research method of diagnostic tests,150 pieces of multi-spectral fundus images demonstrated to be retinal arteriosclerosis of different grades possibly and calibrated by experts were used as the reference standard for labeling in reading,The diagnostic consistency and reading speed among the AI group,the senior ophthalmologist group,junior ophthalmologist group and the cardiologist group were compared.Results As for the diagnostic consistency,the Kendall coordination coefficients were 0.887,0.853 and 0.848(P<0.01)when the AI group was compared with the senior ophthalmologist group,the junior ophthalmologist group and the cardiologist group respectively,so the diagnosis level was close;The average reading time of 1 piece of AI group,the senior ophthalmologist group,the junior ophthalmologist group and the cardiologist group were(1.52±0.29),(14.70±1.74),(22.36±2.43)and(43.10±7.08)s,and when the average reading time of 1 piece were compared among the4 groups,the difference was statistically significant(F=3134.857,P<0.01).Conclusion The combination of AI and multi-spectral fundus imaging technology can improve the screening efficiency of retinal arteriosclerosis,reduce the development difficulties and help doctors of different qualifications,especially general practitioners and young doctors,to quickly grasp the diagnostic and screening methods of this disease.

关 键 词：多光谱眼底成像 人工智能 视网膜动脉硬化 

分 类 号：R816.97[医药卫生—放射医学]