【摘要】 目的 开发一种基于数据驱动的冠状动脉微循环阻力快速计算方法。方法 构建和优化神经网络对冠状动脉进行截面积特征提取，利用截面积特征、异速标度律和流量分配比例快速计算冠状动脉分支末端的微循环阻力，并基于微循环阻力无创计算血流储备分数。结果 为了验证神经网络的有效性，将40个临床收集的冠状动脉分支测量的截面积特征与神经网络预测的结果进行比较，平均绝对误差为1.08 mm~2。为了验证微循环阻力值的准确性，将15位患者的临床血流储备分数与利用微循环阻力值计算的血流储备分数进行比较，计算准确性为86.6%。结论 本文提出的冠状动脉微循环阻力快速计算方法具有潜在的临床应用价值。更多还原

【Abstract】 Objective To developed a data-driven method for fast calculation of coronary microcirculation resistance. Methods The neural network was constructed and optimized to extract cross-sectional area features of coronary arteries. The microcirculation resistance at the end of the coronary branch was quickly calculated by using cross-sectional area features, allometric scaling law and flow distribution ratio, and the blood flow reserve fraction was non-invasively calculated based on microcirculation resistance. Results In order to verify validity of the neural network, the cross-sectional area characteristics of 40 clinically collected coronary artery branch measurements were compared with predicted result of the neural network, and the mean absolute error value was 1.08 mm~2. In order to verify accuracy of the microcirculation resistance, the clinical fractional flow reserve of 15 patients was compared with the fractional flow reserve calculated by the microcirculation resistance, and the calculation accuracy was 86.6%. Conclusions The rapid calculation method of coronary microcirculation resistance proposed in this study has potential clinical application value.更多还原