Gallone Guglielmo, Iodice Francesco, Presta Alberto, Tore Davide, De Filippo Ovidio, Visciano Michele, Barbano Carlo Alberto, Serafini Alessandro, Gorrini Paola, Bruno Alessandro, Grosso Marra Walter, Hughes James, Iannaccone Mario, Fonio Paolo, Fiandrotti Attilio, Depaoli Alessandro, Grangetto Marco, De Ferrari Gaetano Maria, D'Ascenzo Fabrizio
Division of Cardiology, Cardiovascular and Thoracic Department, Città della Salute e della Scienza Hospital, Corso Bramante 88/90, 10126, Turin, Italy.
Department of Medical Sciences, University of Turin, Via Giuseppe Verdi, 8, 10124 Turin, Italy.
Eur Heart J Digit Health. 2025 Apr 21;6(4):567-576. doi: 10.1093/ehjdh/ztaf033. eCollection 2025 Jul.
To develop a deep-learning-based system for recognition of subclinical atherosclerosis on a plain frontal chest X-ray.
A deep-learning algorithm to predict coronary artery calcium (CAC) score (the AI-CAC model) was developed on 460 chest X-ray (80% training cohort, 20% internal validation cohort) of primary prevention patients [58.4% male, median age 63 (51-74) years] with available paired chest X-ray and chest computed tomography (CT) indicated for any clinical reason and performed within 3 months. The CAC score calculated on chest CT was used as ground truth. The model was validated on a temporally independent validation cohort of 90 patients from the same institution (external validation). The diagnostic accuracy of the AI-CAC model assessed by the area under the curve (AUC) was the primary outcome. Overall, median AI-CAC score was 35 (0-388) and 28.9% patients had no AI-CAC. AUC of the AI-CAC model to identify a CAC >0 was 0.90 (95%CI 0.84-0.97) in the internal validation cohort and 0.77 (95%CI 0.67-0.86) in the external validation cohort. Sensitivity was consistently above 92% in both cohorts. In the overall cohort ( = 540), among patients with AI-CAC = 0, a single ASCVD event occurred, after 4.3 years. Patients with AI-CAC > 0 had significantly higher Kaplan Meier estimates for ASCVD events (13.5% vs. 3.4%, log-rank = 0.013).
The AI-CAC model seems to accurately detect subclinical atherosclerosis on chest X-ray with high sensitivity, and to predict ASCVD events with high negative predictive value. Adoption of the AI-CAC model to refine CV risk stratification or as an opportunistic screening tool requires prospective evaluation.
开发一种基于深度学习的系统，用于在胸部正位X线片上识别亚临床动脉粥样硬化。
在460例有配对胸部X线片和胸部计算机断层扫描（CT）的一级预防患者[男性占58.4%，年龄中位数63（51 - 74）岁]的胸部X线片（80%为训练队列，20%为内部验证队列）上开发了一种预测冠状动脉钙化（CAC）评分的深度学习算法（AI - CAC模型），这些患者因任何临床原因进行了胸部X线片和胸部CT检查，且检查在3个月内完成。胸部CT计算出的CAC评分用作金标准。该模型在来自同一机构的90例患者的时间独立验证队列（外部验证）上进行验证。通过曲线下面积（AUC）评估的AI - CAC模型的诊断准确性是主要结局。总体而言，AI - CAC评分中位数为35（0 - 388），28.9%的患者无AI - CAC。AI - CAC模型识别CAC>0的AUC在内部验证队列中为0.90（95%CI 0.84 - 0.97），在外部验证队列中为0.77（95%CI 0.67 - 0.86）。两个队列中的敏感性均始终高于92%。在整个队列（n = 540）中，AI - CAC = 0的患者在4.3年后发生了1例动脉粥样硬化性心血管疾病（ASCVD）事件。AI - CAC>0的患者发生ASCVD事件的Kaplan Meier估计值显著更高（13.5%对3.4%，对数秩检验P = 0.013）。
AI - CAC模型似乎能以高敏感性准确检测胸部X线片上的亚临床动脉粥样硬化，并以高阴性预测值预测ASCVD事件。采用AI - CAC模型来完善心血管风险分层或作为机会性筛查工具需要进行前瞻性评估。
