Hagopian Raffi, Strebel Timothy, Bernatz Simon, Myers Gregory A, Offerman Erik, Zuniga Eric, Kim Cy Y, Ng Angie T, Iwaz James A, Nürnberg Leonard, Singh Sunny P, Carey Evan P, Kim Michael J, Schaefer R Spencer, Yu Jeannie, Gentili Amilcare, Aerts Hugo J W L
Division of Cardiology, Veterans Affairs Long Beach Healthcare System, Long Beach, CA.
Applied Innovations and Medical Informatics (AIMI), Veterans Affairs Long Beach Healthcare System, Long Beach, CA.
NEJM AI. 2025 May 16;2(6). doi: 10.1056/aioa2400937.
Coronary artery calcium (CAC) is highly predictive of cardiovascular events. Although millions of chest computed tomography (CT) scans are performed annually in the United States, CAC is not routinely quantified from scans done for noncardiac purposes.
We developed a deep learning algorithm, AI-CAC, using 446 expert segmentations to automatically quantify CAC on noncontrast, nongated CT scans. Our study differs from prior works by utilizing imaging data from 98 medical centers across the Veterans Affairs national health care system, capturing extensive heterogeneity in imaging protocols, scanners, and patients. AI-CAC performance on nongated scans was compared against clinical standard electrocardiogram (ECG)-gated CAC scoring in 795 patients with paired gated scans within 1 year of their nongated scan. In addition, the model was tested on 8052 low-dose CTs (LDCTs) to simulate opportunistic CAC screening.
Nongated AI-CAC differentiated zero versus nonzero and less than 100 versus 100 or greater Agatston scores with accuracies of 89.4% (F1 0.93) and 87.3% (F1 0.89), respectively. Nongated AI-CAC was predictive of 10-year all-cause mortality (CAC 0 vs. >400 group: 25.4% vs. 60.2%, Cox hazard ratio 3.49; P<0.005), and composite first-time stroke, myocardial infarction, or death (CAC 0 vs. >400 group: 33.5% vs. 63.8%, Cox hazard ratio 3.00; P<0.005). In the LDCT dataset, 3091 out of 8052 (38.4%) individuals had AI-CAC scores >400. Four cardiologists qualitatively reviewed a random sample of the >400 AI-CAC LDCT patients and verified that 527 of the 531 (99.2%) would benefit from lipid-lowering therapy.
This nongated CT CAC algorithm was developed across a national health care system and shows strong performance in evaluation against paired gated CT scans. The model code and weights are available at https://github.com/Raffi-Hagopian/AI-CAC/. (Funded by the Veterans Affairs health care system.).
冠状动脉钙化（CAC）对心血管事件具有高度预测性。尽管美国每年进行数百万次胸部计算机断层扫描（CT），但对于非心脏检查目的的扫描，CAC并未常规进行量化。
我们开发了一种深度学习算法AI-CAC，使用446个专家分割结果在非增强、非门控CT扫描上自动量化CAC。我们的研究与之前的研究不同之处在于，利用了退伍军人事务部国家医疗保健系统中98个医疗中心的成像数据，涵盖了成像协议、扫描仪和患者方面的广泛异质性。在795例在非门控扫描后1年内进行了配对门控扫描的患者中，将非门控扫描上AI-CAC的性能与临床标准心电图（ECG）门控CAC评分进行了比较。此外，该模型在8052例低剂量CT（LDCT）上进行了测试，以模拟机会性CAC筛查。
非门控AI-CAC区分Agatston评分为零与非零以及小于100与100及以上的准确率分别为89.4%（F1值为0.93）和87.3%（F1值为0.89）。非门控AI-CAC可预测10年全因死亡率（CAC为0与>400组：25.4%与60.2%，Cox风险比3.49；P<0.005），以及首次中风、心肌梗死或死亡的复合终点（CAC为0与>400组：33.5%与63.8%，Cox风险比3.00；P<0.005）。在LDCT数据集中，8052例个体中有3091例（38.4%）的AI-CAC评分>400。四位心脏病专家对随机抽取的AI-CAC评分>400的LDCT患者样本进行了定性评估，证实531例中的527例（99.2%）将从降脂治疗中获益。
这种非门控CT CAC算法是在全国医疗保健系统中开发的，在与配对门控CT扫描的比较评估中表现出强大的性能。模型代码和权重可在https://github.com/Raffi-Hagopian/AI-CAC/获取。（由退伍军人事务部医疗保健系统资助。）
