人工智能冠状动脉周围脂肪参数测量的多中心研究

Multi-Center Study on Measurement of Peri-Coronary Adipose Tissue by Artificial Intelligence

目的应用人工智能(AI)测定行冠状动脉CT血管造影(CCTA)的冠状动脉周围脂肪组织(PCAT)脂肪衰减指数(FAI)及体积,探究冠状动脉粥样硬化斑块性质及狭窄程度与PCAT相关性,斑块性质不同时PCAT的FAI在糖尿病、高血压及血脂异常组中的表达是否有特异性。方法回顾性搜集云南省11家综合医院行CCTA检查的患者资料共3561例CTA图像,对所有研究对象的住院记录进行系统回顾,临床资料及实验室检查[包括性别、年龄、体质指数(BMI)、吸烟习惯、饮酒习惯、高血压、糖尿病、血脂异常、脑梗史、心梗史、空腹血糖、糖化血红蛋白]齐全的资料。分析评价斑块性质及冠状动脉狭窄程度,并将原始图像推送至数坤造影图像AI分析软件自动勾画PCAT范围并自动计算三支冠状动脉FAI及PCAT体积。研究不同斑块性质及狭窄程度之间PCAT相关变量是否存在差异,对糖尿病、高血压、血脂异常组冠状动脉斑块性质与FAI、PCAT体积、血脂的进行相关性检验。结果三支冠状动脉PCTA的FAI值、体积与冠状动脉狭窄程度无相关性(P>0.05),而左前降支动脉(LAD)、左旋支动脉(LCX)、右冠状动脉(RCA)的FAI值、RCA-PCAT体积与斑块性质存在相关性(P<0.05)。糖尿病组RCA-FAI与斑块性质存在相关性(P=0.003),高血压组LAD、LCX、RCA的FAI值与斑块性质存在相关性(P=0.003、<0.01、0.010),血脂异常组LAD、LCX的FAI值与斑块性质存在相关性(P=0.010、0.041)。糖尿病组RCA的FAI值在非钙化斑块与钙化斑块间存在统计学差异(P=0.003);高血压组LAD、LCX、RCA的FAI值在非钙化斑块与钙化斑块间均存在统计学差异(P<0.05),LCX、RCA的FAI值在非钙化斑块与混合斑块间存在统计学差异(P<0.05),血脂异常组LAD的FAI值在非钙化斑块与钙化斑块间、RCA的FAI值在非钙化斑块与混合斑块间存在统计学差异(P=0.023、0.011)。糖尿病组、高血压组及血脂异常组中三支冠状动脉的FAI值在钙化斑块与混合斑块间均无统计学差异(P>0.05)。结论利用AI可以快速识别PCAT范围并自动计算相关参数,PCAT的FAI值、体积与冠状动脉狭窄程度无明显相关性,但与冠状动脉斑块性质有相关性。糖尿病组、高血压组及血脂异常症组中PCAT的FAI值的表现是不同的,可以根据不同的基础病结合FAI的异常评估冠状动脉炎性病变。AI测量PCAT快速、简便,对冠状动脉炎症可进行无创性的测量,具有潜在的临床应用价值。

Objective The fat attenuation index(FAI)and volume of peri coronary adipose tissue(PCAT)were measured by artificial intelligence,to study the correlation between coronary plaque and PCAT,the correlation between the degree of coronary stenosis and PCAT.The value of FAI in diabetes,hypertension and dyslipidemia group was different from coronary plaque.Methods In this retrospective and observation study,the images of 3561 patients who underwent coronary CCTA in 11 general hospitals in Yunnan Province were collected.The hospitalization records of all subjects were systematically reviewed,and the clinical data and laboratory examination were collected,the degree of coronary plaque and coronary stenosis were evaluated,then push the image to AI analysis software,automatically outline the PCAT range and automatically calculate FAI and PCAT volume.To explore whether there were differences in PCAT between coronary plaque and stenosis,and to study the correlation between coronary atherosclerotic plaques and PCAT-FAI,volume and blood lipids in diabetes,hypertension and dyslipidemia group were examined.Results There was no correlation between the FAI value and volume of three coronary artery PCTA and the degree of coronary artery stenosis(P>0.05).The FAI of LAD,LCX,RCA and PCAT volume of RCA were correlated with different coronary plaque(P<0.05).There was a correlation between RCA-FAI and plaque character in diabetes group(P=0.003).The FAI of LAD,LCX and RCA in hypertension group were correlated with the nature of plaque(P=0.003,<0.001 and 0.010),the FAI of LAD and LCX in dyslipidemia group were correlated with the nature of plaque(P=0.010,0.041).The FAI of RCA in diabetic group was statistically different between non calcified plaques and calcified plaques(P=0.003).There were significant differences in FAI of LAD,LCX and RCA between non calcified plaque and calcified plaque in hypertension group(P<0.05),there was significant difference in FAI of LCX and RCA between non calcified plaque and mixed plaque(P<0.05).In hyperlipidemia group,the FAI of LAD was significantly different between non calcified plaque and calcified plaque,and the FAI of RCA was significantly different between non calcified plaque and mixed plaque(P=0.023,0.011).The FAI of the three coronary arteries in the diabetic group,the hypertension group and the dyslipidemia group were not statistically different between the calcified plaque and the mixed plaque(P>0.05).Conclusion AI can quickly identify the PCAT range and automatically calculate relevant parameters,the FAI and volume of PCAT had no significant correlation with the degree of coronary artery stenosis,but had correlation with different coronary plaque.The FAI of PCAT in diabetic group,hypertension group and dyslipidemia group are different.They can be used to evaluate coronary artery inflammatory disease based on different basic diseases combined with FAI abnormalities.AI measurement of PCAT is fast and simple,and can measure coronary artery inflammation noninvasively.