深度学习驱动的中心线提取算法对无创冠状动脉血流储备分数“灰区”诊断的优化价值研究Evaluation of a deep learning-driven centerline extraction algorithm for optimizing the diagnosis of the “gray zone” in noninvasive coronary fractional flow reserve
郭自强,王玺,刘子暖,丁熠璞,辛然,单冬凯,郭军,陈韵岱,杨俊杰
摘要(Abstract):
目的 评估基于最小代价路径的CT血流储备分数(MCP-FFR)和深度学习驱动的CT血流储备分数(DeepCL-FFR)的诊断效能,并重点探索DeepCL算法在提升“灰区”诊断准确性方面的潜在价值。方法 收集2020年1月至2021年6月在中国人民解放军总医院第六医学中心住院治疗的109例冠心病患者的151支血管进行回顾性分析。采用Pearson相关性分析和Bland-Altman图评估两种冠状动脉CT血流储备分数(CT-FFR)与FFR的相关性和一致性。将CT-FFR位于0.70~0.80定义为诊断“灰区”。计算并分析诊断血流动力学异常的准确度、敏感度、特异度、阳性预测值和阴性预测值;使用Delong检验比较两种CT-FFR计算方法间受试者工作特征(ROC)的曲线下面积(AUC)。结果 两种CT-FFR与FFR之间均成正相关(MCP-FFR:r=0.75,P <0.001;DeepCL-FFR:r=0.86,P <0.001),且均具有良好的一致性(MCP-FFR:平均值0.010,P=0.351;DeepCL-FFR:平均值–0.003,P=0.772)。DeepCL-FFR(AUC 0.97,95%CI0.94~0.99)与MCP-FFR(AUC0.92,95%CI0.88~0.97)对血流动力学异常均具有较佳的诊断效能(P=0.122)。在“灰区”血流动力学异常的诊断中,MCP-FFR的诊断准确性为68.8%,而DeepCL-FFR则提升至89.7%。DeepCL-FFR在“灰区”血流动力学异常诊断中表现出较佳的诊断效能(AUC 0.89, 95%CI 0.73~0.99),显著高于MCP-FFR(AUC 0.71,95%CI 0.54~0.87)(P<0.001)。结论 深度学习驱动的冠状动脉中心线提取算法DeepCL,在CT-FFR诊断血流动力学异常方面具有较佳的诊断效能,特别是在“灰区”诊断中显著提高了诊断准确性。
关键词(KeyWords): 深度学习;冠心病;血流储备分数;中心线提取;全卷积神经网络
基金项目(Foundation): 国家重点研发计划项目(2021YFC2500505);; 国家自然科学基金项目(82470342);; 北京市科技新星计划项目(20230484471)
作者(Author): 郭自强,王玺,刘子暖,丁熠璞,辛然,单冬凯,郭军,陈韵岱,杨俊杰
参考文献(References):
- [1] Pijls NH, Fearon WF, Tonino PA, et al. Fractional flow reserve versus angiography for guiding percutaneous coronary intervention in patients with multivessel coronary artery disease:2-year follow-up of the FAME(Fractional Flow Reserve Versus Angiography for Multivessel Evaluation)study[J]. J Am Coll Cardiol, 2010,56(3):177-184. DOI:10.1016/j.jacc.2010.04.012.
- [2] Li Y, Qiu H, Hou Z, et al. Additional value of deep learning computed tomographic angiography-based fractional flow reserve in detecting coronary stenosis and predicting outcomes[J]. Acta Radiol, 2022,63(1):133-140. DOI:10.1177/0284185120983977.
- [3] Cook CM, Petraco R, Shun-Shin MJ, et al. Diagnostic accuracy of computed tomography-derived fractional flow reserve:a systematic review[J]. JAMA Cardiol, 2017,2(7):803-810. DOI:10.1001/jamacardio.2017.1314.
- [4] Brandt V, Fischer A, Schoepf UJ, et al. Deep learning-based automated labeling of coronary segments for structured reporting of coronary computed tomography angiography in accordance with society of cardiovascular computed tomography guidelines[J]. J Thorac Imaging, 2024,39(2):93-100. DOI:10.1097/RTI.0000000000000753.
- [5] Zhang Y, Luo G, Wang W, et al. A continuous-action deep reinforcement learning-based agent for coronary artery centerline extraction in coronary CT angiography images[J]. Med Biol Eng Comput, 2025,63(6):1837-1847. DOI:10.1007/s11517-025-03284-3.
- [6] Li R, Davoodi A, Cai Y, et al. Robot-assisted ultrasound reconstruction for spine surgery:from bench-top to pre-clinical study[J]. Int J Comput Assist Radiol Surg, 2023,18(9):1613-1623. DOI:10.1007/s11548-023-02932-z.
- [7] Guo Z,Bai J,Lu Y,et al. DeepCenterline:a multi-task fully convolutional network for centerline extration[C]//Chung A,Gee J,Yushkevich P,Bao S,eds. Information Processing in Medical Imaging. 2PMI 2019:441-453. DOI:10.1007/978-3-030-20351-1-34.
- [8]中国医师协会心血管内科医师分会超声心动图和影像学组,《中国介入心脏病学杂志》编辑委员会.冠状动脉CT血流储备分数应用临床路径中国专家共识[J].中国介入心脏病学杂志,2023,31(4):241-251. DOI:10.3969/j.issn.1004-8812.2023.04.001.
- [9]吴迅,崔建国,金琴花.冠状动脉功能学评估在经皮冠状动脉介入治疗术后的应用进展[J].中国介入心脏病学杂志,2023,31(3):219-224. DOI:10.3969/j.issn.1004-8812.2023.03.009.
- [10] Guo W, He W, Lu Y, et al. CT-FFR by expanding coronary tree with Newton-Krylov-Schwarz method to solve the governing equations of CFD[J]. Eur Heart J Imaging Methods Pract, 2024,2(3):qyae106. DOI:10.1093/ehjimp/qyae106.
- [11] Chu M, Wu P, Li G, et al. Advances in diagnosis, therapy,and prognosis of coronary artery disease powered by deep learning algorithms[J]. JACC Asia, 2023, 3(1):1-14. DOI:10.1016/j.jacasi.2022.12.005.
- [12] Ma Y, Wu L, Gao Y, et al. ULFAC-Net:ultra-lightweight fully asymmetric convolutional network for skin lesion segmentation[J]. IEEE J Biomed Health Inform, 2023,27(6):2886-2897. DOI:10.1109/JBHI.2023.3259802.
- [13] Wolterink JM,van Hamersvelt RW, Viergever MA, et al.Coronary artery centerline extraction in cardiac CT angiography using a CNN-based orientation classifier[J]. Med Image Anal, 2019,51:46-60. DOI:10.1016/J.MEDIA.2018.10.005.
- [14]曹慧峰,朱均超,刘欣宜,等.一种快速光栅条纹中心亚像素坐标提取方法[J].激光杂志,2021, 42(3):52-57.DOI:10.14016/j.cnki.jgzz.2021.03.052.
- [15]《中国冠状动脉血流储备分数测定技术临床路径专家共识》专家组.中国冠状动脉血流储备分数测定技术临床路径专家共识[J].中国介入心脏病学杂志,2019,27(3):121-133.DOI:10.3969/j.issn.1004-8812.2019.03.001.
- [16] Zimmermann FM,Omerovic E,Fournier S,et al. Fractional flow reserve-guided percutaneous coronary intervention vs.medical therapy for patients with stable coronary lesions:metaanalysis of individual patient data[J]. Eur Heart J,2019,40(2):180-186. DOI:10.1093/eurheartj/ehy812.