摘要
目的:评估基于胸部低剂量CT的人工智能(AI)骨密度测量系统与定量CT (QCT)测定骨密度的相关性和一致性,并评估其诊断骨质疏松症的有效性。方法:回顾性纳入1 427名接受了低剂量胸部CT检查的健康体检者,分别使用AI骨密度测量系统和QCT测量T12、L1、L2椎体骨密度。采用线性回归分析和Bland-Altman法评价2种方法测定骨密度的相关性和一致性。以QCT结果为参考标准,绘制受试者工作特征(ROC)曲线评估AI骨密度测量系统诊断骨质疏松的准确性。结果:2种方法测量T12、L1和L2骨密度的相关系数分别为0.988、0.992、0.990;2种方法测量的骨密度差值分别为2.96 (95%CI 2.62~3.31)、1.46(95%CI 1.18~1.74)、1.61 (95%CI 1.30~1.92)mg/cm3,大部分位于均值±1.96SD范围内。AI系统预测骨密度用于诊断骨质疏松的ROC曲线下面积为0.996 (95%CI 0.994~0.998;P<0.001),灵敏度为0.989,特异度为0.905。结论:AI骨密度测量系统与QCT测定骨密度具有高度相关性及一致性,可有效诊断骨质疏松症。
Purpose:To evaluate the correlation and agreement between the developed artificial intelligence(AI)system based on low-dose chest CT images and quantitative computed tomography(QCT)for measuring bone mineral density(BMD),and to test the AI system for diagnosing osteoporosis.Methods:1427 healthy volunteers who underwent low-dose chest CT examination were included,and the BMDs were measured at the T12,L1,L2 vertebral bodies using the AI bone densitometry system and QCT,respectively.The linear regression and Bland-Altman analysis were used to evaluate the correlation and consistency of the two methods.The efficacy of the AI system for diagnosing osteoporosis was assessed using receiver operating characteristic(ROC)curve analysis,and the QCT results were the golden standard.Results:The correlation coefficients between the AI bone densitometry system and QCT measurements were 0.988,0.992 and 0.990 for T12,L1 and L2,respectively.The Bland-Altman analysis showed that the differences of BMD between the two methods were 2.96(95%CI 2.62-3.31),1.46(95%CI 1.18-1.74)and 1.61(95%CI 1.30-1.92)mg/cm3,respectively,and most of which fell within the range of mean±1.96SD.The area under the ROC curve of the AI system for the diagnosis of osteoporosis using QCT as the diagnostic criterion was 0.996(95%CI 0.994-0.998;P<0.001),with a sensitivity of 0.989 and a specificity of 0.905.Conclusions:The AI bone densitometry system is highly correlated and consistent with QCT in determining bone density,and can effectively diagnose osteoporosis.
作者
陈刘萍
余卓
潘亚玲
王晗琦
陆勇
CHEN Liuping;YU Zhuo;PAN Yaling;WANG Hanqi;LU Yong(Department of Radiology,Ruijin Hospital,Shanghai Jiao Tong University School of Medicine;Huiying Medical Technology(Beijing)Co.;Department of Radiology,The People′s Hospital of Zhejiang Province)
出处
《中国医学计算机成像杂志》
CSCD
北大核心
2023年第2期178-183,共6页
Chinese Computed Medical Imaging
基金
上海市自然科学基金(21ZR1439800)。
关键词
定量CT
骨密度
骨质疏松
人工智能
Quantitative computed tomography
Bone mineral density
Osteoporosis
Artificial intelligence