| 孟颖,王志远,张纪,沈龙山,王震寰,陈刘成.基于人工智能冠状动脉CT血管造影多参数特征联合临床指标预测斑块进展[J].中国医学影像技术,2025,41(9):1506~1511 |
| 基于人工智能冠状动脉CT血管造影多参数特征联合临床指标预测斑块进展 |
| Multi-parameter coronary CT angiography features based on artificial intelligence combined with clinical indicators for predicting plaque progression |
| 投稿时间：2024-12-31 修订日期：2025-07-17 |
| DOI：10.13929/j.issn.1003-3289.2025.09.011 |
| 中文关键词: 冠状动脉疾病|动脉粥样硬化|疾病进展|CT血管造影|人工智能 |
| 英文关键词:coronary artery disease|atherosclerosis|disease progression|computed tomography angiography|artificial intelligence |
| 基金项目:数字医学与智慧健康安徽省重点实验室(蚌埠医科大学)开放课题基金项目(AHCM2024Z005)、蚌埠医科大学2024年度研究生科研创新计划项目(Byycx24103)。 |
| 作者 | 单位 | E-mail | | 孟颖 | 蚌埠医科大学第一附属医院放射科, 安徽 蚌埠 233004 | | | 王志远 | 蚌埠医科大学第一附属医院放射科, 安徽 蚌埠 233004 | | | 张纪 | 蚌埠医科大学第一附属医院放射科, 安徽 蚌埠 233004 | | | 沈龙山 | 数字医学与智慧健康安徽省重点实验室, 蚌埠医科大学临床解剖应用研究所, 安徽 蚌埠 233004
蚌埠医科大学第二附属医院放射科, 安徽 蚌埠 233040 | | | 王震寰 | 数字医学与智慧健康安徽省重点实验室, 蚌埠医科大学临床解剖应用研究所, 安徽 蚌埠 233004 | | | 陈刘成 | 蚌埠医科大学第一附属医院放射科, 安徽 蚌埠 233004
数字医学与智慧健康安徽省重点实验室, 蚌埠医科大学临床解剖应用研究所, 安徽 蚌埠 233004 | chenliu1385521@163.com |
|
| 摘要点击次数: 135 |
| 全文下载次数: 44 |
| 中文摘要: |
| 目的 观察基于人工智能(AI)的冠状动脉CT血管造影(CCTA)多参数特征联合临床指标预测冠状动脉斑块进展的价值。方法 回顾性分析143例冠状动脉动脉粥样硬化(AS)患者,以斑块负荷算术平均年增长率＞1%或≤1%将其分为进展组(n=73)与未进展组(n=70);比较2组基线临床资料、CT血流储备分数(CT-FFR)、冠状动脉周围脂肪衰减指数(FAI)及斑块定量特征。针对组间有统计学差异的变量剔除其中与其他变量存在明显共线性者,利用多因素logistic回归筛选斑块进展的独立预测因子并构建联合模型;绘制受试者工作特征(ROC)曲线,以曲线下面积(AUC)评估联合模型预测效能。结果 进展组合并高血压、糖尿病者占比,以及载脂蛋白A1(ApoA1)和高敏C反应蛋白(hs-CRP)均高于,而高密度脂蛋白胆固醇(HDL-C)低于非进展组(P均＜0.05)。进展组最小管腔面积及CT-FFR小于,而管腔最狭窄程度、斑块总体积、斑块负荷、非钙化斑块体积、脂质斑块体积、纤维脂质斑块体积及FAI均大于非进展组(P均＜0.05);组间斑块类型亦存在差异(P＜0.05)。合并糖尿病、HDL-C低、最小管腔面积小及脂质斑块体积大均为冠状动脉AS患者斑块进展的独立预测因子(P均＜0.05),以之构建的联合模型预测斑块进展的AUC为0.859。结论 基于AI的CCTA多参数特征联合临床指标可有效预测冠状动脉AS斑块进展。 |
| 英文摘要: |
| Objective To explore the value of artificial intelligence (AI) based multi-parameter coronary CT angiography (CCTA) features combined with clinical indicators for predicting coronary plaque progression. Methods Totally 143 coronary atherosclerosis (AS) patients were retrospectively enrolled and divided into progression group (arithmetic average annual growth rate of plaque load＞1%, n=73) and non-progression group (arithmetic average annual growth rate of plaque load≤1%, n=70). The baseline clinical data, CT-derived fractional flow reserve (CT-FFR), perivascular fat attenuation index (FAI), and quantitative plaque features were collected and compared between groups. For variables being statistically different between groups, those had collinearity with others were excluded, and then multivariable logistic regression was used to screen independent predictors of plaque progression from the retained variables, and a combined model was constructed. Receiver operating characteristic (ROC) curve was drawn, and the area under the curve (AUC) was calculated to evaluate the predictive efficacy of this model. Results Progression group had higher proportions of hypertension and diabetes, higher apolipoprotein A1 (ApoA1) and high-sensitivity C-reactive protein (hs-CRP) levels but lower high-density lipoprotein cholesterol (HDL-C) levels than non-progression group (all P＜0.05). Progression group showed smaller minimum lumen area and lower CT-FFR, but greater degree of lumen stenosis, total plaque volume, plaque load, non-calcified plaque volume, lipid-rich plaque volume, fibrolipid plaque volume and FAI values than non-progression group (all P＜0.05). Plaque types were different between groups (P＜0.05). Diabetes, low HDL-C, small minimum lumen area and large lipid-rich plaque volume were all independent predictors of plaque progression in patients with coronary AS (all P＜0.05), and the AUC of the combined model for predicting plaque progression was 0.859. Conclusion Multi-parameter CCTA features based on AI combined with clinical indicators could be used to effectively predict progression of coronary AS plaque. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
