Association between Non-Stenosis Carotid Plaque Characteristics and Embolic Stroke of Undetermined Source in Patients with Unilateral Anterior Circulation Infarction: A CTA Analysis Based AI
Objective: To evaluate the association between characteristics of non-carotid plaque and embolic stroke of undetermined source (ESUS) on computed tomography angiography (CTA) in patients with unilateral anterior circulation infarction with artificial intelligence (AI). Methods: Patients with acute unilateral anterior circulation ischemic stroke admitted to the hospital from May 2022 to March 2024 were identified and included as large artery atherosclerosis stroke (LAAS), ESUS, and cardioembolic stroke (CES). The clinical features of ESUS, LAAS, and CES patients were compared and AI-based on CTA images automatically analyzed and compared the differences in ipsilateral and contralateral carotid plaque features of ESUS patients with ischemic events and explored their correlation. Results: We analyzed 72 patients with LAAS, 50 patients with ESUS, and 30 patients with CES and found that ESUS patients had younger onset age and lower C-reactive protein levels than LAAS patients. ESUS patients had more vascular risk factors and fewer abnormal cardiac markers than CES patients. In terms of imaging, ESUS patients had more mixed or non-calcified plaques in the ipsilateral carotid artery than the contralateral carotid artery, and the incidence of mixed or non-calcified plaques in lumen stenosis was higher. Logistic regression analysis showed that both were independent influencing factors of ESUS. Comparing the features of carotid plaque on the side of LAAS and ESUS ischemic events, it was found that mixed or non-calcified carotid plaque of LAAS was more common. Conclusion: The clinical risk factors for ESUS are closer to LAAS than CES, and mixed or non-calcified plaques are more common on the same side of ischemic events, which may be a potential etiological marker for ESUS.
Embolic Stroke of Undetermined Source
回顾性纳入2022年5月至2024年3月期间青岛大学附属医院平度院区神经内科收治的单侧前循环脑梗死患者。纳入标准:符合LAAS
收集患者基线资料,包括人口统计学特征(年龄、性别)、血管危险因素(高血压、糖尿病、冠心病、吸烟史、饮酒史)、基线血压(收缩压、舒张压)、实验室检查(全血白细胞计数、血小板计数、D-二聚体、C反应蛋白、总胆固醇、甘油三酯、低密度脂蛋白胆固醇(Low Density Lipoprotein Cholesterol, LDL)、高密度脂蛋白胆固醇(High Density Lipoprotein Cholesterol, HDL)、脂蛋白a、空腹血糖、糖化血红蛋白、高同型半胱氨酸、尿酸、血清钙、血清磷)、心脏功能相关指标(脑利钠肽、左房前后径、左心室舒张EDT、左心室舒张功能E'e)、美国国立卫生研究院卒中量表(National Institutes of Health Stroke Scale, NIHSS)评分、出院时改良Rankin量表(modified Rankin Scale, mRS)评分。
选取CTA作为影像评估的主要方式,将CTA影像上传至数坤科技平台用于评估LAAS和ESUS患者颈动脉狭窄程度及颈动脉斑块的特征。数坤软件是一款已通过国家药品监督管理局批准的商业AI软件,其产品训练是通过人工标注勾画病灶区域,机器学习大量病例的影像特征后而自动分析颈动脉各段的狭窄程度、斑块的位置及特征(钙化斑块、混合斑块、非钙化斑块),示例分析结果见
采用SPSS 26.0软件进行统计学分析。符合正态分布的计量资料以均数 ± 标准差表示,不符合正态分布的计量资料以中位数和四分位数间距表示,两组正态分布、方差齐性比较采用独立样本t检验,两组非正态分布或方差不齐者比较采用Mann Whitney U检验。计数资料以频数和百分率表示,组间比较采用χ2检验。缺血性事件同侧和对侧颈动脉特征比较若为定量资料且差值正态分布采用配对的学生t检验,若差值不符合正态分布则采用符号秩和检验;定性变量如同侧和对侧斑块特征的发生率比较采用McNemar检验。应用二分类多因素Logistic回归分析确定与ESUS及LAAS患者有关的斑块影响因素,并计算优势比(Odds Ratio, OR)和95%置信区间(Confidence Interval, CI),P值 < 0.05认为有显著性差异。
变量 |
LAAS组 |
ESUS组 |
P值 |
人口统计学 |
|||
年龄(岁) |
70 (63.25~77.75) |
64.42 ± 11.67 |
0.024 |
男性(n, %) |
55 (76.4) |
38 (76) |
0.960 |
NIHSS评分 |
1.5 (1~3.75) |
1 (0.0~3.0) |
0.253 |
mRS评分 |
1 (1~2) |
1 (1.0~1.0) |
0.058 |
血管危险因素 |
|||
高血压(n, %) |
51 (70.8) |
33 (66) |
0.571 |
糖尿病(n, %) |
16 (22.2) |
15 (30) |
0.332 |
冠心病(n, %) |
8 (11.1) |
3 (6) |
0.522 |
脑梗死(n, %) |
22 (30.6) |
12 (24) |
0.427 |
吸烟史(n, %) |
31 (43.1) |
25 (50) |
0.449 |
饮酒史(n, %) |
19 (26.4) |
20 (40) |
0.113 |
基线收缩压(mmHg) |
156.57 ± 23.85 |
149.92 ± 21.07 |
0.115 |
基线舒张压(mmHg) |
85.17 ± 13.98 |
85.78 ± 12.42 |
0.804 |
实验室检查 |
|||
白细胞计数(109/L) |
7.22 (6.27~8.54) |
6.97 ± 1.50 |
0.096 |
血小板计数(109/L) |
241.03 ± 73.63 |
248.56 ± 66.66 |
0.565 |
D-二聚体(ng/mL) |
370 (295.0~537.5) |
330 (270~400) |
0.056 |
C反应蛋白(mg/L) |
2.82 (0.84~7.77) |
1.4 (0.76~2.41) |
0.020 |
总胆固醇(mmol/L) |
4.21 ± 1.21 |
4.03 ± 0.92 |
0.362 |
甘油三酯(mmol/L) |
1.25 (0.91~1.57) |
1.16 (0.79~1.55) |
0.255 |
脂蛋白a (mg/L) |
206.15 (98.43~433.93) |
164.1 (92.73~280.28) |
0.274 |
LDL (mmol/L) |
2.81 ± 1.06 |
2.56 ± 0.78 |
0.150 |
HDL (mmol/L) |
1.05 ± 0.24 |
1.07 ± 0.21 |
0.675 |
空腹血糖(mmol/L) |
5.18 (4.73~5.92) |
4.96 (4.59~6.38) |
0.500 |
糖化血红蛋白 |
6 (5.55~7.0) |
5.8 (5.3~6.7) |
0.173 |
同型半胱氨酸(umol/L) |
12.6 (10.05~17.0) |
12.2 (9.88~15.83) |
0.608 |
尿酸(umol/L) |
290.27 ± 80.42 |
295.5 (258~352) |
0.234 |
血清钙(mmol/L) |
2.20 ± 0.13 |
2.20 ± 0.88 |
0.790 |
血清磷(mmol/L) |
1.13 ± 0.18 |
1.18 ± 0.14 |
0.198 |
心脏相关指标 |
|||
脑利钠肽(pg/mL) |
127.0 (53.15~302.25) |
75.9 (36.05~155.0) |
0.073 |
左房前后径(cm) |
3.84 ± 0.35 |
3.79 ± 0.27 |
0.432 |
左房扩大(n, %) |
34 (73.9) |
28 (71.8) |
0.827 |
左心室射血分数(%) |
60 (60.0~61.0) |
60 (60.0~62.0) |
0.101 |
左心室舒张EDT (ms) |
219.91 ± 29.47 |
212.93 ± 30.57 |
0.357 |
左心室舒张功能E'e |
12.1 (11.1~13.3) |
11.2 (10.6~12.2) |
0.054 |
变量 |
CES组 |
ESUS组 |
P值 |
人口统计学 |
|||
年龄(岁) |
67.97 ± 11.20 |
64.42 ± 11.67 |
0.186 |
男性(n, %) |
17 (56.7) |
38 (76) |
0.071 |
NIHSS评分 |
2 (0.75~3) |
1 (0.0~3.0) |
0.537 |
mRS评分 |
1 (0~1) |
1 (1.0~1.0) |
0.485 |
血管危险因素 |
|||
高血压(n, %) |
15 (50) |
33 (66) |
0.157 |
糖尿病(n, %) |
9 (30) |
15 (30) |
1.000 |
冠心病(n, %) |
3 (10) |
3 (6) |
0.667 |
脑梗死(n, %) |
5 (16.7) |
12 (24) |
0.438 |
吸烟史(n, %) |
5 (16.7) |
25 (50) |
0.003 |
饮酒史(n, %) |
4 (13.3) |
20 (40) |
0.012 |
基线收缩压(mmHg) |
135.83 ± 20.63 |
149.92 ± 21.07 |
0.005 |
基线舒张压(mmHg) |
82 (74.75~101.5) |
85.78 ± 12.42 |
0.881 |
实验室检查 |
|||
白细胞计数(109/L) |
7.21 ± 2.16 |
6.97 ± 1.50 |
0.562 |
血小板计数(109/L) |
213 (167~251.25) |
248.56 ± 66.66 |
0.03 |
D-二聚体(ng/mL) |
355 (287.5~502.5) |
330 (270~400) |
0.239 |
C反应蛋白(mg/L) |
1.25 (0.6~4.30) |
1.4 (0.76~2.41) |
0.886 |
总胆固醇(mmol/L) |
4.44 ± 0.91 |
4.03 ± 0.92 |
0.050 |
甘油三酯(mmol/L) |
1.11 (0.75~1.66) |
1.16 (0.79~1.55) |
0.952 |
脂蛋白a (mg/L) |
213.55 (113.3~339.78) |
164.1 (92.73~280.28) |
0.185 |
LDL (mmol/L) |
2.88 ± 0.85 |
2.56 ± 0.78 |
0.088 |
HDL (mmol/L) |
1.18 ± 0.24 |
1.07 ± 0.21 |
0.031 |
空腹血糖(mmol/L) |
5.57 (4.84~6.01) |
4.96 (4.59~6.38) |
0.311 |
糖化血红蛋白 |
6.1 (5.5~7.5) |
5.8 (5.3~6.7) |
0.188 |
同型半胱氨酸(umol/L) |
9.7 (8.78~11.23) |
12.2 (9.88~15.83) |
0.003 |
尿酸(umol/L) |
294 (233.5~330.5) |
295.5 (258~352) |
0.411 |
血清钙(mmol/L) |
2.21 ± 0.11 |
2.20 ± 0.88 |
0.717 |
血清磷(mmol/L) |
1.13 ± 0.19 |
1.18 ± 0.14 |
0.273 |
心脏相关指标 |
|||
脑利钠肽(pg/mL) |
1160 (485.5~1830) |
75.9 (36.05~155.0) |
<0.001 |
左房前后径(cm) |
4.35 (3.93~4.98) |
3.79 ± 0.27 |
<0.001 |
左房扩大(n, %) |
27 (96.4) |
28 (71.8) |
0.009 |
左心室射血分数(%) |
58 (55.25~60) |
60 (60.0~62.0) |
<0.001 |
左心室舒张EDT (ms) |
192.22 ± 37.07 |
212.93 ± 30.57 |
0.100 |
左心室舒张功能E'e |
16.06 ± 7.27 |
11.2 (10.6~12.2) |
0.024 |
采用配对统计方法比较ESUS组患者缺血事件同侧及对侧颈动脉狭窄程度、斑块数量、颈动脉最狭窄处不同性质斑块发生率的差异。考虑到既往研究发现颈动脉斑块多集中于颈动脉分叉处上方2厘米~下方2厘米,我们对双侧颅外颈动脉(颈总动脉、颈内动脉C1段)狭窄程度及斑块性质进行了比较。结果提示ESUS组患者缺血事件同侧颈动脉混合或非钙化斑块数量较对侧更多,颈动脉最狭窄处混合或非钙化斑块发生率更高,结果存在统计学差异,见
变量 |
ESUS (n = 33) |
P值 |
|
同侧 |
对侧 |
||
颈动脉狭窄程度(%) |
35 (14~41.50) |
30 (18~60.0) |
0.137 |
颅外段颈动脉段狭窄程度(%) |
23 (2~34) |
13 (0~23.50) |
0.755 |
颈动脉斑块数量(个) |
4 (2~5) |
3 (1~6) |
0.848 |
颅外段颈动脉斑块数量(个) |
2 (1~3) |
1 (0~2.5) |
0.369 |
混合或非钙化斑块数量(个) |
1 (0~1) |
0 (0~1) |
0.033 |
斑块发生率(n, %) |
32 (64) |
31 (62) |
1.000 |
钙化斑块发生率(n, %) |
25 (75.8) |
30 (90.9) |
0.180 |
混合斑块发生率(n, %) |
16 (48.5) |
8 (24.2) |
0.057 |
非钙化斑块发生率(n, %) |
9 (27.3) |
6 (18.2) |
0.549 |
混合或非钙化斑块发生率(n, %) |
22 (66.7) |
14 (42.4) |
0.077 |
颅外颈动脉最狭窄处混合或非钙化斑块发生率(n, %) |
20 (74.1) |
12 (52.2) |
0.180 |
颈动脉最狭窄处混合或非钙化斑块发生率(n, %) |
21 (65.6) |
9 (28.1) |
0.002 |
变量 |
OR |
95% CI |
P值 |
混合或非钙化斑块数量 |
2.545 |
1.082~5.986 |
0.032 |
颈动脉最狭窄处混合或非钙化斑块发生率 |
5.469 |
1.883~15.884 |
0.002 |
变量 |
LAAS组 |
ESUS组 |
P值 |
狭窄程度(%) |
63.60 ± 31.40 |
35 (14~41.50) |
<0.001 |
颈动脉斑块数量(个) |
4 (3~5) |
4 (2~5) |
0.206 |
颅外段颈动脉斑块数量(个) |
2 (1~3) |
2 (1~3) |
0.800 |
混合或非钙化斑块数量(个) |
1 (1~1) |
1 (0~1) |
0.008 |
钙化斑块发生率(n, %) |
54 (90.0) |
25 (78.1) |
0.131 |
混合斑块发生率(n, %) |
32 (53.3) |
16 (50.0) |
0.760 |
非钙化斑块发生率(n, %) |
27 (45.0) |
9 (28.1) |
0.114 |
混合或非钙化斑块发生率(n, %) |
56 (93.3) |
22 (68.8) |
0.004 |
颅外颈动脉最狭窄处混合或非钙化斑块发生率(n, %) |
52 (92.9) |
20 (74.1) |
0.034 |
颈动脉最狭窄处混合或非钙化斑块发生率(n, %) |
41 (68.3) |
21 (65.6) |
0.792 |
研究表明超声和MRI上颈动脉斑块的某些形态学特征是卒中的独立危险因素,且不管狭窄的程度如何。对比高分辨MRI较高的设备标准及价格要求,CTA更容易获得且作为大多数中心急性中风成像方案的一部分常规进行。因此在CTA上识别和评估非狭窄性颈动脉斑块的特征将在临床中将更加实用。CTA可提供丰富的病理生理信息,包括管腔狭窄程度、斑块溃疡、斑块钙化程度等,但在临床实践中由于大量图像需要人工解读,加之高级成像技术依赖后期处理,传统CT图像分析耗时较长,故仍以主观判读和常规征象识别为主,缺乏对图像信息的进一步识别以及客观定量定性处理,对于<50%狭窄的颈动脉斑块的成像特征知之甚少。有研究表明,CTA上的非狭窄性颈动脉斑块与同侧卒中相关,但未发现高危的CTA斑块特征。随着AI的发展,以及机器学习在医学影像分析领域的应用,其在头颈部动脉狭窄患者的诊断和治疗中,尤其是急性缺血性脑梗死方面已展现出新的前景
本研究比较了三组患者的临床特征,结果显示,ESUS组与LAAS组患者在年龄及CRP水平存在统计学差异。ESUS组患者发病年龄更小,这与既往研究表明ESUS更常见于年轻患者相一致
ESUS的临床危险因素更接近LAAS而不是CES。在ESUS患者中,缺血事件同侧的混合或非钙化颈动脉斑块数量更占优势、颈动脉最狭窄处混合或非钙化斑块发生率更高,并与卒中风险独立相关。AI辅助的CTA分析可为ESUS病因筛查提供实用工具,但其临床价值需通过前瞻性研究进一步验证。
*通讯作者。