基于虚拟组织学-血管内超声的人工智能辅助冠状动脉CT血管成像斑块定量分析系统开发与验证Development and validation of an artificial intelligence-powered coronary CT angiography analysis system using virtual histology intravascular ultrasound-derived radiomics for coronary plaque characterization
兰永昊,孙泽宇,殷维钧,蔡濛,杨文涛,李嘉辉,刘巍
摘要(Abstract):
目的 构建基于虚拟组织学-血管内超声(VH-IVUS)的冠状动脉CT血管成像(CCTA)人工智能分析系统,以VH-IVUS为金标准验证人工智能解析冠状动脉管腔及斑块成分的能力。方法 连续筛选2023年1月至2024年12月在首都医科大学附属北京积水潭医院接受冠状动脉造影联合IVUS检查,且术前1个月内行CCTA的患者。收集105例患者188支病变血管的曲面重建图像(CPR)与IVUS影像(均符合质量标准)。医师手动标注VH-IVUS病灶范围,分割为坏死核心、纤维组织、纤维脂质、致密钙4种(颜色标记),并标注冠状动脉开口用于配准。CCTA图像经血管分割后生成拉直CPR图,人工标注开口位置。采用卷积神经网络(CNN)结合循环神经网络(RNN)混合模型提取特征并回归斑块成分占比,以验证集性能为评价标准。按80%:20%划分训练集(84例/150支血管)与独立测试集(21例/38支血管)。模型经PyTorch框架训练[均方误差(MSE)损失函数,Adam优化器,初始学习率0.01,权重衰减0.001,余弦退火策略,200轮次],最终实现CPR斑块与管腔特征可视化。结果 模型斑块定量预测平均MSE=0.0018。各成分Pearson相关系数均>0.730(P<0.0001):纤维组织(r=0.795)、坏死核心(r=0.782)、致密钙(r=0.762)、纤维脂质(r=0.730)。R~2系数:纤维组织0.628,余成分0.451~0.628。Bland-Altman分析示>95%数据点位于95%一致性界限内,纤维组织离散度最低(SD=0.048)。结论 融合三维CNN与RNN人工智能系统可从CCTA中提取VH-IVUS级腔内信息,实现高精度斑块表征。
关键词(KeyWords): 血管内超声;冠状动脉CT血管成像;深度学习;斑块成分分析;卷积神经网络
基金项目(Foundation): 北京市自然科学基金面上项目(7232078);; 北京积水潭医院院级科研基金项目(QN202213)
作者(Author): 兰永昊,孙泽宇,殷维钧,蔡濛,杨文涛,李嘉辉,刘巍
参考文献(References):
- [1]Knuuti J,Wijns W,Saraste A,et al.2019 ESC Guidelines for the diagnosis and management of chronic coronary syndromes[J].Eur Heart J,2020,41(3):407-477.DOI:10.1093/eurheartj/ehz425.
- [2]陈韵岱.人工智能赋能心血管学科发展[J].中国介入心脏病学杂志,2025,33(6):301-302.DOI:10.3969/j.issn.1004-8812.2025.06.001.
- [3]陈韵岱.聚焦高水平人工智能研究,优化心血管介入精准施治[J].中国介入心脏病学杂志,2024,32(1):1-4.DOI:10.3969/j.issn.1004-8812.2024.01.001.
- [4]血管内超声在冠状动脉疾病中应用的中国专家共识专家组.血管内超声在冠状动脉疾病中应用的中国专家共识(2018)[J].中华心血管病杂志,2018,46(5):344-351.DOI:10.3760/cma.j.issn.0253-3758.2018.05.005.
- [5]Goeller M,Achenbach S,Duncker H,et al.Imaging of the pericoronary adipose tissue(PCAT)using cardiac computed tomography:modern clinical implications[J].J Thorac Imaging,2021,36(3):149-161.DOI:10.1097/RTI.0000000000000583.
- [6]Yuvaraj J,Cheng K,Lin A,et al.The emerging role of ct-based imaging in adipose tissue and coronary inflammation[J].Cells,2021,10(5):1196.DOI:10.3390/cells10051196.
- [7]Honold S,Wildauer M,Beyer C,et al.Reciprocal communication of pericoronary adipose tissue and coronaryatherogenesis[J].Eur J Radiol,2021,136:109531.DOI:10.1016/j.ejrad.2021.1095 3 1.
- [8] Oikonomou EK,Marwan M,Desai MY,et al.Non-invasive detection of coronary inflammation using computed tomography and prediction of residual cardiovascular risk(the CRISP CT study):a post-hoc analysis of prospective outcome data[J].Lancet,2018,392(10151):929-939.DOI:10.1016/S0140-6736(18)31114-0.
- [9] Oikonomou EK,Desai MY,Marwan M,et al.Perivascular fat attenuation index stratifies cardiac risk associated with high-risk plaques in the CRISP-CT study[J].J Am Coll Cardiol,2020,76(6):755-757.DOI:10.1016/j.jacc.2020.05.078.
- [10]Nair A,Kuban BD,Tuzcu EM,et al.Coronary plaque classification with intravascular ultrasound radiofrequency data analysis[J].Circulation,2002,106(17):2200-2206.DOI:10.1161/01.cir.0000035654.18341.5e.
- [11] Nasu K,Tsuchikane E,Katoh O,et al.Accuracy of in vivo coronary plaque morphology assessment:a validation study of in vivo virtual histology compared with in vitro histopathology[J].J Am Coll Cardiol,2006,47(12):2405-2412.DOI:10.1016/j.jacc.2006.02.044.
- [12]Zhang X,Broersen A,Sokooti H,et al.Cross-sectional angle prediction of lipid-rich and calcified tissue on computed tomography angiography images[J].Int J Comput Assist Radiol Surg,2024,19(5):971-981.DOI:10.1007/s 11548-024-03086-2.
- [13]Hochreiter S,Schmidhuber J.Long short-term memory[J].Neural Comput,1997,9(8):1735-1780.DOI:10.1162/neco.1997.9.8.1735.
- [14]Narula J,Stuckey TD,Nakazawa G,et al.Prospective deep learning-based quantitative assessment of coronary plaque by computed tomography angiography compared with intravascular ultrasound:the REVEALPLAQUE study[J].Eur Heart J Cardiovasc Imaging,2024,25(9):1287-1295.DOI:10.1093/ehjci/jeae115.
- [15]Ramasamy A,Hamid A Khan A,Cooper J,et al.Implications of computed tomography reconstruction algorithms on coronary atheroma quantification:comparison with intravascular ultrasound[J].J Cardiovasc Comput Tomogr,2023,17(1):43-51.DOI:10.1016/j.jcct.2022.09.004.
- [16]Jin Q,Fu Z,Wang Y,et al.A multicenter feasibility and safety study of a novel hybrid IVUS-OCT imaging system[J].JACC Asia,2025,5(3 Pt 1):396-400.DOI:10.1016/j.jacasi.2025.01.009.