冠脉CT影像处理软件中的AI算法性能测试概述  

Overview of AI Algorithm Performance Testing in Coronary CT Image Processing Software

作　　者：王晨希[1] 王浩[1] 李佳戈[1] WANG Chenxi;WANG Hao;LI Jiage(Division of Active Medical Device and Medical Optics,National Institutes for Food and Drug Control,Beijing 100050,China)

机构地区：[1]中国食品药品检定研究院光机电室,北京100050

出　　处：《中国医疗设备》2023年第4期31-34,共4页China Medical Devices

基　　金：工信部“面向医疗健康行业的人工智能筛查和辅助诊断公共服务平台”项目(2020-0103-3-1)。

摘　　要：近年来,冠脉CT影像处理软件陆续被批准上市,此类产品的风险点在于人工智能算法的性能评价,本文根据冠脉CT影像处理软件产品的实际情况,提出了测试环境、数据集要求、测试平台、测试指标和通过准则等方面的测试要求,针对分段识别和检出、分类和分割功能提出测试方法和相应的评价指标,指出产品输出不同或测试集参考标准不同,可能导致评价方法不同,从而影响到测试结果。In recent years,Coronary CT Image Processing Software has been approved to market successively.The risk of such products lies in the performance evaluation of artificial intelligence algorithm.According to the actual situation of coronary CT image processing software products,this paper put forward test requirements in terms of test environment,data set requirements,test platform,test indicators and pass criteria,and put forward test methods and corresponding evaluations for segment recognition and detection,classification and segmentation functions,It was concluded that different product outputs or test set reference standards may lead to different evaluation methods,thus affecting the discussion of test results.

关 键 词：人工智能 算法 测试集 方法 评价 

分 类 号：R197.39[医药卫生—卫生事业管理]