中文摘要

第一部分  正常冠状动脉中管腔直径变化对CT-FFR的影响

【背景】 最近报道，即使在非阻塞性冠状动脉疾病（NOCAD）背景下，计算机断层扫描衍生的血流储备分数（CT-FFR）也表现出从冠状动脉近段到远段逐渐下降趋势，这导致约14.1%的NOCAD患者血管远段CT-FFR出现假阳性诊断。目前尚不清楚CT-FFR的进行性下降是由于狭窄本身亦或是其它潜在因素影响的结果。

【目的】 本研究旨在探讨正常冠状动脉中CT-FFR的变化趋势，以及管腔直径自然变化与CT-FFR变化趋势之间的潜在联系。

【材料及方法】 回顾性连续收集2022年9月1日至2022年9月30日期间于中国医学科学院阜外医院就诊且行冠状动脉CT血管成像（CCTA）检查的门诊患者2200例。纳入标准为年龄 ≥ 18岁且CCTA检查显示所有冠状动脉正常，不存在任何性质的粥样硬化斑块。所有患者均进行CCTA图像分析及CT-FFR分析，从冠状动脉开口起以每10mm间隔单位测量三支冠状动脉的CT-FFR值和管腔直径值，直至远段血管管腔直径 < 1.5mm停止测量，绘制CT-FFR和管腔直径变化曲线图；并计算每10mm单位距离内CT-FFR和管腔直径的变化（ΔCT-FFR、ΔLD）、CT-FFR衰减梯度（CDG）和管腔直径递减梯度（TDG）。使用Spearman相关分析和线性回归分析CT-FFR衰减梯度和管腔直径递减梯度之间的联系。

【结果】 研究最终纳入了177名患者（年龄49.2 ± 9.6岁，46.3%男性），529支冠状动脉（177支前降支、175支回旋支和177支右冠状动脉）。三支冠状动脉的CT-FFR均显示出沿血管长轴逐渐衰减改变。其中，前将支表现出最明显的下降，在血管末端的CT-FFR值最低（前降支：0.839 ± 0.052；回旋支：0.896 ± 0.052；右冠状动脉：0.865 ± 0.044）。此外，前降支和右冠状动脉表现出CT-FFR双相下降模式，即靠近血管开口和血管远段出现快速下降期，但回旋支只观察到血管远段一个快速下降期。Spearman分析表明CDG和TDG之间呈正相关（前降支：r = 0.432，P < 0.001；回旋支：r = 0.477，P < 0.001；右冠状动脉：r = 0.488，P < 0.001）。三支冠状动脉的CDG和TDG线性回归方程分别为 y_前降支 = 0.023x + 0.033（P < 0.001）、y_回旋支 = 0.029x + 0.024（P < 0.001）和 y_{右冠状动脉} = 0.033x + 0.038（P < 0.001）。

【结论】 CT-FFR显示出与狭窄无关的自然衰减，并且CT-FFR衰减梯度与管腔直径递减梯度呈线性正相关；基于CT-FFR的临床决策应考虑其生理衰减特性，尤其是对于远段病变。

第二部分  不同测量位置对计算流体力学CT-FFR诊断缺血病变效能的影响：CT-FFR CHINA临床试验结果

【背景】 无论是否存在狭窄，CT血流储备分数（CT-FFR）都显示出从血管开口到远段逐渐衰减趋势。

【目的】 探讨不同测量位置对CT-FFR测量值及诊断效能的影响；并以压力导丝FFR为标准确定最佳CT-FFR测量位置。

【材料及方法】 CT-FFR CHINA临床试验连续入选2018年11月至2020年3月，于全国5家临床机构就诊且临床疑诊冠心病（CAD）的410例患者。所有患者均在7天内依次完成冠状动脉CT血管成像（CCTA）、CT-FFR、侵入性冠状动脉造影（ICA）和FFR检查。纳入标准为年龄 ≥ 18岁且CCTA检查发现冠状动脉至少存在一处狭窄 ≥ 30% 且 ≤ 90%。将纳入患者的目标病变血管按不同CT-FFR测量位置分为五组：病变结束处、病变后1cm、病变后2cm、病变后3cm及血管末端组。以侵入性FFR作为诊断缺血病变的标准，采用Spearman及Bland-Altman法比较不同测量位置组CT-FFR测量值与FFR值的相关性及一致性；并评估其对缺血病变的诊断性能，包括准确性、特异性、敏感性、阴性预测值、阳性预测值和受试者工作特征曲线下面积（AUC），确定最佳CT-FFR测量位置。

【结果】 此研究纳入296例患者（68.9%为男性，平均年龄59.0 ± 9.5岁），包含352条目标病变血管。CT-FFR与FFR测量差值的绝对值随测量距离越远呈现逐渐增大的趋势[病变结束处组：0.010（95% LoA：-0.203 ~ 0.223），病变后1cm组：0.014（-0.199 ~ 0.228），病变后2cm组：0.027（-0.182 ~ 0.236），病变后3cm组：0.040（-0.167 ~ 0.247），血管末端组：0.075（-0.135 ~ 0.285）]。在血管水平，病变后1cm的CT-FFR诊断心肌缺血的准确性[84.66%（95%CI: 80.46% ~ 88.26%）]最高，且不同测量位置组间差异具有统计学意义（P = 0.008）。此外，病变后1cm组AUC效能[0.85（95%CI: 0.81 ~ 0.89）]最高，与病变结束处组[0.83（0.79 ~ 0.87），P = 0.028]、病变后2cm组[0.82（0.77 ~ 0.86），P = 0.001]、病变后3cm组[0.79（0.74 ~ 0.83），P < 0.001]、血管末端组[0.77（0.72 ~ 0.81），P < 0.001]组间比较均具有统计学差异。

【结论】 CT-FFR测量值受测量位置影响，CT-FFR的测量误差呈现随测量位置越远而增大的趋势；目标病变后1cm测量CT-FFR是识别CAD患者心肌缺血的最佳选择。

第三部分  CT-FFR对不同冠脉节段病变的诊断效能差异：CT-FFR CHINA临床试验结果

【背景】 无论是否存在狭窄，CT血流储备分数（CT-FFR）都显示出从血管开口到远段逐渐衰减趋势。

【目的】 本研究的目的是讨论计算流体力学CT-FFR对于不同冠脉节段病变心肌缺血诊断性能的差异。

【材料及方法】 CHINA CT-FFR临床试验连续入选2018年11月至2020年3月，于全国5家临床机构就诊且临床疑诊冠心病（CAD）的410例患者。所有患者均在7天内依次完成冠状动脉CT血管成像（CCTA）、无创CT-FFR、侵入性冠状动脉造影（ICA）和有创FFR检查。纳入标准为年龄 ≥ 18岁且CCTA检查发现冠状动脉至少存在一处狭窄 ≥30%且≤ 90%。在血管水平，根据冠状动脉十八节段划分法，将目标病变血管按不同病变节段部位分为四组：近段、中段、远段、多节段组。以压力导丝FFR作为缺血病变的诊断标准，计算并比较不同病变节段组的CT-FFR测量值误差，以及诊断缺血病变的性能[包括准确性、敏感性、特异性、阴性预测值、阳性预测值和受试者工作特征曲线下面积（AUC）]。

【结果】 此研究纳入296例患者（68.9%为男性，平均年龄59.0 ± 9.5岁），包含352条目标病变血管。在血管水平，所有目标病变的CT-FFR测量值和FFR值高度相关（r = 0.736，P < 0.001）；但随着病变所处冠脉节段越远，两者的相关性呈现减弱趋势。以压力导丝FFR为标准，CT-FFR对冠状动脉缺血病变的诊断准确性、敏感性、特异性和阴性预测值在不同病变节段组没有显著差异，但远段组的阳性预测值显著低于其它组[近段组：84.62%（95%CI: 74.19% ~ 91.32%）；中段组：73.08%（62.86% ~ 81.32%）；远段组：47.06%（31.55% ~ 63.16%）；多节段组：85.71%（77.44% ~ 91.29%）；P = 0.006]。相较于近段病变，CT-FFR对中段及远段病变的诊断准确性（90.65% vs. 80.58% vs. 78.85%，P = 0.113）、敏感性（95.65% vs. 86.36% vs. 80.00%，P = 0.198）、AUC效能（0.91 vs. 0.81 vs. 0.79）逐渐降低。其中，近段组与远段组的AUC存在统计学差异[0.91（95%CI: 0.84 ~ 0.96）vs. 0.79（0.66 ~ 0.89），P = 0.033]。

【结论】 CT-FFR测量值与FFR值相关性呈现随病变节段越远而逐渐减弱趋势；不同病变节段的差异并没有显著影响CT-FFR诊断冠状动脉缺血病变的准确性；但相较于近段病变，远段病变的阳性预测值和AUC效能显著减低。

第四部分  基于计算流体力学CT-FFR诊断心肌缺血的优化研究：CT-FFR、校正CT-FFR、ΔCT-FFR三种方法比较

【背景】 无论是否存在狭窄，CT血流储备分数（CT-FFR）都显示出从血管开口到远段逐渐衰减趋势。

【目的】 探讨计算流体力学CT-血流储备分数（CT-FFR）、经函数校正CT-FFR（校正CT-FFR）、跨病变两端CT-FFR差值（ΔCT-FFR）对冠状动脉缺血病变的诊断价值。

【材料及方法】 CHINA CT-FFR临床试验连续入选2018年11月至2020年3月于全国5家临床机构就诊且临床疑诊冠心病（CAD）的410例患者。所有入选者均在7天内依次完成冠状动脉CT血管成像（CCTA）、无创CT-FFR、侵入性冠状动脉造影（ICA）和有创FFR检查。纳入标准为年龄 ≥ 18岁且CCTA检查发现冠状动脉至少存在一处狭窄 ≥ 30% 且 ≤ 90%。在血管水平，分析目标病变的CT-FFR、校正CT-FFR和ΔCT-FFR值。采用Spearman相关分析评估三种方法与侵入性FFR之间的相关性。以侵入性FFR ≤ 0.8作为缺血病变的诊断标准，比较三种无创血流动力学方法识别缺血病变的诊断性能[包括准确性、特异性、敏感性、阴性预测值（NPV）、阳性预测值（PPV）和受试者工作特征曲线下面积（AUC）]。将上述结果中最佳无创血流动力学方法添加进CCTA狭窄程度（DS%）评估模型，使用AUC评估各模型预测心肌缺血的效能，使用净重分类指数（NRI）和综合判别指数（IDI）对新旧模型进行比较，分析无创血流动力学评估方法对于预测心肌缺血的增量价值。

【结果】 在血管水平，三种方法与侵入性FFR均具有高度相关性（r = 0.736，r = 0.716，r = -0.718；all P < 0.001）。与非缺血病变（FFR > 0.8）相比，缺血病变（FFR ≤ 0.8）显示出更高的CT-FFR值（0.68 ± 0.13 vs. 0.85 ± 0.08，P < 0.001）、校正CT-FFR值（0.75 ± 0.12 vs. 0.92 ± 0.08，P < 0.001）和ΔCT-FFR值（0.27 ± 0.12 vs. 0.10 ± 0.08，P < 0.001）。三种方法中，ΔCT-FFR显示出最高的诊断准确性[86.36%（95%CI: 82.33% ~ 89.77%）]及AUC效能[0.87（95%CI: 0.83 ~ 0.90）]；其中，ΔCT-FFR的AUC与校正CT-FFR[0.80（95%CI: 0.76 ~ 0.84），P = 0.002]存在统计学差异，但与CT-FFR[0.85（95%CI: 0.81 ~ 0.89），P = 0.317]未观察到显著差异。将无创血流动力学检测参数（CT-FFR和ΔCT-FFR）添加进CCTA狭窄程度模型后，包含血流动力学信息的模型2（DS% + CT-FFR）和模型3（DS% + CT-FFR + ΔCT-FFR）预测心肌缺血的性能明显提高；其中，包含ΔCT-FFR的模型3比模型2显示出更好的区分度和重分类能力[AUC: 0.90（0.86 ~ 0.93），P = 0.003；NRI: 7.5%（1.7% ~ 13.3%），P = 0.001；IDI: 5.5%（3.3% ~ 7.8%），P < 0.001]。

【结论】 三种无创血流动力学方法均与侵入性FFR密切相关，但ΔCT-FFR显示出最高的诊断准确性及AUC效能，对预测心肌缺血具有增量价值；将ΔCT-FFR纳入临床CCTA和CT-FFR评估流程中，可能会改变患者的治疗策略、减少不必要的侵入性ICA检查。

Abstract

Part 1  Computed tomography derived fractional flow reserve variation in normal coronary arteries

【Background】 Recent studies reported that computed tomography (CT) derived fractional flow reserve (CT-FFR) may gradually decline from the ostium to distal segment vessels even in case of non-obstructive coronary artery disease (NOCAD). The result is that approximately 14.1% of NOCAD patients receive false positive diagnoses on distal vessel CT-FFR. However, it is unclear whether the progressive decrease in CT-FFR is due to the stenosis itself or other potential influencing factors.

【Objective】 This study aimed to describe the changing trend in CT-FFR along three coronary arteries, and to investigate the impact of luminal diameter on CT-FFR variations, as well as establish relationships between natural variations in luminal diameter and corresponding changes observed in CT-FFR.

【Materials and Methods】 This study retrospectively analyzed 2200 outpatients who underwent coronary CT angiography (CCTA) at Fuwai hospital during September 2022. Inclusion criteria were age ≥ 18 years and CCTA examination showed that all coronary arteries were normal without plaques. CCTA image analysis and CT-FFR analysis were performed on all patients. CT-FFR and luminal diameter were measured at 10 mm intervals from the vessel ostium to the distal segment. The CT-FFR and luminal diameter changing trend curves were plotted. Additionally, alterations in CT-FFR and luminal diameter within a unit distance (ΔCT-FFR, ΔLD), CT-FFR decremental gradient (CDG), and transluminal diameter decremental gradient (TDG) were calculated. The correlation between CDG and TDG was determined using Spearman correlation and linear regression analysis.

【Results】 A total of 177 patients (mean age 49.2 ± 9.6 years, 46.3% male) with 529 normal coronary arteries (177 LADs, 175 LCXs, and 177 RCAs) were included. CT-FFR values of three coronary arteries gradually decreased along the vessels, similar to the luminal diameter decrease pattern. LADs showed the greatest decline at terminal vessels (LADs: 0.839 ± 0.052, LCXs: 0.896 ± 0.052, and RCAs: 0.865 ± 0.044). Additionally, LADs and RCAs exhibit a biphasic decline in CT-FFR, with rapid decreases near vessel ostium and distal segment, while LCXs only observed one significant decrease in distal segment. Spearman analysis demonstrated a moderate correlation between CDG and TDG (LADs: r = 0.432, LCXs: r = 0.477, and RCAs: r = 0.488, all P < 0.001). Linear regression equations of CDG and TDG for three coronary arteries were y_(LADs) = 0.023x + 0.033, y_(LCXs) = 0.029x + 0.024, and y_(RCAs) = 0.033x + 0.038, respectively (all P < 0.001).

【Conclusions】 CT-FFR demonstrates a natural decline pattern independent of stenosis, and CDG is proportional to TDG. Decision-making based on CT-FFR value should consider inherent decline characteristics in clinical practice, particularly in distal segment lesions.

Part 2  Effect of measurement site on diagnostic performance of CT-FFR with computational fluid dynamics: results from CT-FFR CHINA clinical trial

【Background】 Regardless of the presence of stenosis, computed tomography derived fractional flow reserve (CT-FFR) shows a gradual decline trend from the ostium to terminal vessel.

【Objective】To investigate the effect of measurement site on the values and diagnostic performance of CT-FFR, and to determine the optimal measurement site for CT-FFR using pressure wire assessment as a reference.

【Materials and Methods】 In CT-FFR CHINA clinical trial, 410 individuals with suspected coronary artery disease were enrolled. Each participant underwent coronary CT angiography and CT-FFR, followed by invasive coronary angiography (ICA) and invasive FFR examinations within a week of each other. The target lesion vessels were classified into five groups based on different CT-FFR measurement sites: post-lesion, 1 cm, 2 cm, 3 cm distal to the lesion, and terminal vessel groups. The CT-FFR values for measurement deviation were compared. Diagnostic accuracy and performance [including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the receiver operating characteristic curve (AUC)] of CT-FFR in discriminating myocardial ischemia were analyzed across all measurement site groups on a per-vessel level, using FFR as the reference standard.

【Results】 A total of 296 patients (mean age of 59.0 ± 9.5 years) with 352 target lesion vessels were included in this study. The absolute difference between CT-FFR and FFR measurements increased as the measurement site distance increased: post-lesion group, 0.010 (95%LoA: -0.203 ~ 0.223), CT-FFR_1cm group, 0.014 (-0.199 ~ 0.228), CT-FFR_2cm group, 0.027 (-0.182 ~ 0.236), CT-FFR_3cm group, 0.040 (-0.167 ~ 0.247), and terminal vessel group, 0.075 (-0.135 ~ 0.285). At a per-vessel level, the accuracy of diagnosing myocardial ischemia with CT-FFR_1cm was the highest at 84.66% (95% CI: 80.46% ~ 88.26%), with statistically significant differences between measurement site groups (P = 0.008). The AUC of the CT-FFR_1cm group [0.85 (95%CI: 0.81 ~ 0.89)] was also the highest, showing significant differences compared to the post-lesion group [0.83 (0.79 ~ 0.87), P = 0.028], CT-FFR_2cm group [0.82 (0.77 ~ 0.86), P = 0.001], CT-FFR_3cm group [0.79 (0.74 ~ 0.83), P < 0.001], and terminal vessel group [0.77 (0.72 ~ 0.81), P < 0.001].

【Conclusions】 The deviation of CT-FFR increases with measurement site distance. CT-FFR of 1 cm distal to the target lesion is the optimal choice for diagnosing myocardial ischemia in patients with coronary artery disease.

Part 3  Variations in CT-FFR diagnostic performance for different coronary lesion segment: results from CT-FFR CHINA clinical trial

【Background】 Regardless of the presence of stenosis, computed tomography derived fractional flow reserve (CT-FFR) shows a gradual decline trend from the ostium to terminal vessel.

【Objective】 This study aimed to discuss the diagnostic performance differences of computational fluid dynamics CT-FFR in diagnosing myocardial ischemia across different coronary segment lesions.

【Materials and Methods】 In CHINA CT-FFR clinical trial, 410 individuals with suspected coronary artery disease were enrolled. Each participant underwent coronary CT angiography and CT-FFR, followed by invasive coronary angiography (ICA) and invasive FFR examinations within a week of each other. Using the eighteen-segment division method of coronary arteries, based on target lesion segment location, the target lesion vessels were classified into four groups: proximal segment, middle segment, distal segment, and multi-segment groups. Using invasive FFR as the standard for diagnosing myocardial ischemia, CT-FFR measurement deviation were compared. Diagnostic accuracy and performance [including sensitivity, specificity, positive predictive value, negative predictive value, and area under the receiver operating characteristic curve (AUC)] of CT-FFR were analyzed across all lesion segment location groups on a per-vessel level.

【Results】 A total of 296 patients (mean age of 59.0 ± 9.5 years) with 352 target lesion vessels were included in this study. Across all target lesions, CT-FFR values were highly correlated with invasive FFR (r = 0.737, P < 0.001), however, the correlation coefficient decreased with increasing distance of lesion segment location. The accuracy, sensitivity, specificity, and negative predictive value of CT-FFR for discriminating myocardial ischemia between different lesion segments are not significantly different. However, the positive predictive value of the distal segment group was significantly lower than other groups [proximal segment group: 84.62% (95%CI: 74.19% ~ 91.32%), middle segment group: 73.08% (62.86% ~ 81.32%), distal segment group: 47.06% (31.55%~63.16%), multi-segment group: 85.71% (77.44% ~ 91.29%), P = 0.006]. Compared with proximal segment lesions, CT-FFR diagnostic accuracy (90.65% vs. 80.58% vs. 78.85%, P = 0.113), sensitivity (95.65% vs. 86.36% vs. 80.00%, P = 0.198), and AUC (0.91 vs. 0.81 vs. 0.79) gradually decreased in middle and distal segment lesions. Among them, the AUC [0.91 (95%CI: 0.84 ~ 0.96) vs. 0.79 (0.66 ~ 0.89), P = 0.033] showed statistical differences between the proximal and the distal segment group.

【Conclusions】 The correlation between CT-FFR and FFR gradually decreases with increased lesion segment location. While the lesion segment does not significantly affect CT-FFR diagnostic accuracy, the positive predictive value and AUC of the distal segment group are significantly lower than proximal segment group.

Part 4  Optimized methods of CT-FFR with computational fluid dynamics for discriminating myocardial ischemia: comparison of three interpretation methods

【Background】 Regardless of the presence of stenosis, computed tomography derived fractional flow reserve (CT-FFR) shows a gradual decline trend from the ostium to terminal vessel.

【Objective】 This study aimed to discuss the performance of three CT-FFR measurements in detecting myocardial ischemia by comparing them with invasive FFR, including CT-FFR, corrected CT-FFR and ΔCT-FFR.

【Materials and Methods】 In CHINA CT-FFR clinical trial, 410 individuals with suspected coronary artery disease were enrolled. Each participant underwent coronary CT angiography and CT-FFR, followed by invasive coronary angiography (ICA) and invasive FFR examinations within a week of each other. The correlation between three methods and FFR was determined using Spearman correlation analysis. The diagnostic accuracy and performance [including sensitivity, specificity, positive predictive value, negative predictive value, and area under the receiver operating characteristic curve (AUC)] of three methods in diagnosing myocardial ischemia were analyzed, using FFR as the reference standard. According to the above results, the most effective method was added to the CCTA stenosis degree (DS%) model. To assess the incremental discrimination of ischemia, an AUC analysis was performed, and the reclassification performance of each model was compared using relative integrated discrimination improvement (IDI) index and the category-free net reclassification index (NRI).

【Results】 At a per-vessel level, all three methods were highly correlated with invasive FFR (r = 0.736, r = 0.716, r = -0.718, all P < 0.001). Compared to non-ischemic lesions (FFR > 0.8), ischemic lesions (FFR ≤ 0.8) showed higher CT-FFR values, corrected CT-FFR values, and ΔCT-FFR values. Among the three methods, ΔCT-FFR showed the highest diagnostic accuracy and AUC, the optimal cut-off value was 0.149. A statistically significant difference existed between ΔCT-FFR and corrected CT-FFR, but not between ΔCT-FFR and CT-FFR. The addition of hemodynamic information improved the discrimination and reclassification abilities of the assessment ischemia compared to the CCTA stenosis evaluation alone, and the addition of information about ΔCT-FFR ≥ 0.149 further increased the discrimination [AUC: 0.90 (95%CI: 0.86 ~ 0.93), P = 0.003] and reclassification abilities [NRI: 7.5% (95%CI: 1.7% ~ 13.3%), P = 0.001；IDI: 5.5% (95%CI: 3.3% ~ 7.8%), P < 0.001] of the assessments.

【Conclusions】 The three methods are closely related to invasive FFR, but ΔCT-FFR shows the highest diagnostic accuracy and AUC efficiency. Additionally, ΔCT-FFR is an incremental predictor of myocardial ischemia, adding ΔCT-FFR to CCTA+CT-FFR assessment may change patient treatment strategies, reducing unnecessary ICA examinations.