基于机器学习的脑卒中高危人群无症状性颈动脉狭窄的检测研究

doi:10.24920/003703

摘要/Abstract

摘要：

目的无症状性颈动脉狭窄（asymptomatic carotid stenosis, ACS）与严重的脑血管病关系密切。早期识别ACS患者及相关危险因素有助于脑卒中的一级预防。本研究旨在探讨一种基于相关危险因素的机器学习算法来检测脑卒中高危人群中ACS发生情况。
方法依据ACS可能的30项危险因素,我们建立一种新的机器学习模型。该模型的算法采用基于训练数据的随机森林模式,然后利用测试数据计算并校准。所有原始数据均来自中国脑卒中筛查与预防项目（China National Stroke Screening and Prevention Project,CNSSPP）。以脑卒中高危个体为研究对象,按照4:1比例随机分为训练组和测试组。根据颈动脉超声检查结果,将训练组分为ACS组和对照组。用受试者工作特性曲线(receiver operating characteristic, ROC)的曲线下面积（area under the curve,AUC）值来验证模型对检测ACS的有效性。
结果本研究共纳入2841例脑卒中高危个体,其中326例（11.6%）确诊为ACS。经过模型计算,引起ACS的前五位危险因素是血脂异常家族史、低密度脂蛋白胆固醇升高、高密度脂蛋白胆固醇降低、高龄、和低体重指数。它们的权重分别为11.8%,7.6%,7.1%,6.1%和6.1%。前15位危险因素的总体权重为85.5%。测试组数据验证本机器学习模型在脑卒中高危人群中检测ACS的AUC值为0.888。
结论本研究证明机器学习算法可用于脑卒中高危人群中ACS的危险因素识别。血脂异常家族史可能是预测ACS发病的首要危险因素。该机器模型可以用作脑卒中一级预防优化临床方法的工具。

关键词: 脑卒中, 高危人群, 无症状性颈动脉狭窄, 危险因素, 机器学习

Abstract:

Objective Asymptomatic carotid stenosis (ACS) is closely associated to the incidence of severe cerebrovascular diseases. Early identifying the individuals with ACS and its associated risk factors could be beneficial for primary prevention of stroke. This study aimed to investigate a machine-learning algorithm for the detection of ACS among high-risk population of stroke based on the associated risk factors.
Methods A novel model of machine learning was utilized to screen the associated predictors of ACS based on 30 potential risk factors. The algorithm of this model adopted a random forest pattern based on the training data and then was verified using the testing data. All of the original data were retrieved from the China National Stroke Screening and Prevention Project (CNSSPP), including demographic, clinical and laboratory characteristics. The individuals with high risk of stroke were enrolled and randomly divided into a training group and a testing group at a ratio of 4:1. The identification of carotid stenosis by carotid artery duplex scans was set as the golden standard. The receiver operating characteristic (ROC) curve and the area under the curve (AUC) was used to evaluate the efficacy of the model in detecting ACS.
Results Of 2841 high risk individual of stroke enrolled, 326 (11.6%) were diagnosed as ACS by ultrasonography. The top five risk factors contributing to ACS in this model were identified as family history of dyslipidemia, high level of low-density lipoprotein cholesterol (LDL-c), low level of high-density lipoprotein cholesterol (HDL-c), aging, and low body mass index (BMI). Their weights were 11.8%, 7.6%, 7.1%, 6.1%, and 6.1%, respectively. The total weight of the top 15 risk factors was 85.5%. The AUC values of the model for detecting ACS with training dataset and testing dataset were 0.927 and 0.888, respectively.
Conclusions This study demonstrated that the machine-learning algorithm could be used to identify the risk factors for ACS among high risk population of stroke. Family history of dyslipidemia may be the most important risk factor for ACS. This model could be a suitable tool to optimize the clinical approach for the primary prevention of stroke.

Key words: high-risk population, stroke, asymptomatic carotid stenosis, risk factors, machine learning

Junxiong Yin, Cheng Yu, Lixia Wei, Chuanyong Yu, Hongxing Liu, Mingyang Du, Feng Sun, Chongjun Wang, Xiaoshan Wang. Detection of Asymptomatic Carotid Artery Stenosis in High-Risk Individuals of Stroke Using a Machine-Learning Algorithm[J].Chinese Medical Sciences Journal, 2020, 35(4): 297-305.

图/表 6

参考文献 29.

1.	DE Waard DD, Morris D, DE Borst GJ, et al. Asym-ptomatic carotid artery stenosis: who should be screened, who should be treated and how should we treat them? J Cardiovasc Surg 2017; 58(1):3-12. doi: 10.23736/S0021-9509.16.09770-6.
2.	Faxon DP, Creager MA, Smith SC, et al. American Heart Association. Atherosclerotic Vascular Disease Conference: Executive summary: Atherosclerotic Vascular Disease Conference proceeding for healthcare professionals from a special writing group of the American Heart Association. Circulation 2004; 109(21):2595-604. doi: 10.1161/01.CIR.0000128517.52533.DB. pmid: 15173041
3.	Fayad ZA, Fuster V. Clinical imaging of the high-risk or vulnerable atherosclerotic plaque. Circ Res 2001; 89(4):305-16. doi: 10.1161/hh1601.095596. pmid: 11509446
4.	U.S. Preventive Services Task Force. Screening for carotid artery stenosis: U.S. Preventive Services Task Force recommendation statement. Ann Intern Med 2007; 147(12):854-9. doi: 10.7326/0003-4819-147-12-200712180-00005. doi: 10.7326/0003-4819-147-12-200712180-00005 pmid: 18087056
5.	Summaries for patients. Screening for carotid artery stenosis: U.S. Preventive Services Task Force recommendation statement. nn Intern Med 2014; 161(5). doi: 10.7326/P14-9030.
6.	Qureshi AI, Alexandrov AV, Tegeler CH, et al; American Society of Neuroimaging; Society of Vascular and Interventional Neurology. Guidelines for screening of extracranial carotid artery disease: a statement for healthcare professionals from the multidisciplinary practice guidelines committee of the American Society of Neuroimaging; cosponsored by the Society of Vascular and Interventional Neurology. J Neuroimaging 2007; 17(1):19-47. doi: 10.1111/j.1552-6569.2006.00085.x. doi: 10.1111/j.1552-6569.2006.00085.x pmid: 17238868
7.	Brott TG, Halperin JL, Abbara S, et al. 2011 ASA/ACCF/AHA/AANN/AANS/ACR/ASNR/CNS/SAIP/SCAI/SIR/SNIS/SVM/SVS Guideline on the management of patients with extracranial carotid and vertebral artery disease. Stroke 2011; 42(8):e464-540. doi: 10.1161/STR.0b013e3182112cc2. Corrected and republished: Stroke 2011; 42(8):e542. Stroke 2012; 43(8):e80. doi: 10.1161/STR.0b013e3182112cc2 pmid: 21282493
8.	Ambale-Venkatesh B, Yang X, Wu CO, et al. Cardiovascular event prediction by machine learning: the multi-ethnic study of atherosclerosis. Circ Res 2017; 121(9):1092-101. doi: 10.1161/CIRCRESAHA.117.311312. doi: 10.1161/CIRCRESAHA.117.311312 pmid: 28794054
9.	Erickson BJ, Korfiatis P, Akkus Z, et al. Machine learning for medical imaging. Radiographics 2017; 37(2):505-15. doi: 10.1148/rg.2017160130. doi: 10.1148/rg.2017160130 pmid: 28212054
10.	Li B, Jiao Y, Fu C, et al. Contralateral artery enlargement predicts carotid plaque progression based on machine learning algorithm models in apoE-/- mice. Biomed Eng Online 2016; 15(Suppl 2):146. doi: 10.1186/s12938-016-0265-z. doi: 10.1186/s12938-016-0265-z pmid: 28155719
11.	Jacobowitz GR, Rockman CB, Gagne PJ, et al. A model for predicting occult carotid artery stenosis: Screening is justified in a selected population. J Vasc Surg 2003; 38(4):705-9. doi: 10.1016/s0741-5214(03)00730-4. doi: 10.1016/s0741-5214(03)00730-4 pmid: 14560217
12.	Greco G, Egorova NN, Moskowitz AJ, et al. A model for predicting the risk of carotid artery disease. Ann Surg 2013; 257(6):1168-73. doi: 10.1097/SLA.0b013e31827b9761. pmid: 23333880
13.	Breiman, Leo. Random forests. Machine learning 2001; 45(1):5-32. Available at https://www.stat.berkeley.edu/~breiman/randomforest2001.pdf. Accessed: December 1, 2019. doi: 10.1023/A:1010933404324
14.	Wang X, Li W, Song F, et al. Carotid atherosclerosis detected by ultrasonography: A national cross-sectional study. J Am Heart Assoc 2018; 7(8):e008701. doi: 10.1161/JAHA.118.008701. doi: 10.1161/JAHA.118.008701 pmid: 29622590
15.	Hua Y, Hui PJ, Xing YQ. Chinese stroke vascular ultrasound examination guidelines. Chin J Med Ultrasound (Electronic Edition) 2015; 12(8):599-610. Chinese. doi: 10.3877/cma.j.issn.1672-6448.2015.08.004.
16.	Touboul PJ, Hennerici MG, Meairs S, et al. Mannheim carotid intima-media thickness consensus (2004-2006). An update on behalf of the Advisory Board of the 3rd and 4th Watching the Risk Symposium, 13th and 15th European Stroke Conferences, Mannheim, Germany, 2004, and Brussels, Belgium, 2006. Cerebrovasc Dis 2007; 23(1):75-80. doi: 10.1159/000097034.
17.	Abbott AL. Medical (nonsurgical) intervention alone is now best for prevention of stroke associated with asymptomatic severe carotid stenosis: results of a systematic review and analysis. Stroke 2009; 40(10):e573-83. doi: 10.1161/STROKEAHA.109.556068. doi: 10.1161/STROKEAHA.109.556068 pmid: 19696421
18.	Abadi M, Barham P, Chen J, et al. Tensorflow: A system for large-scale machine learning. In: OSDI' 16 proceedings of the 12th USENIX Symposium on Operating Systems Design and Implementation. USA: Usenix Association; 2016. p. 265-83.
19.	Jiao Y, Chen R, Ke X, et al. Predictive models of autism spectrum disorder based on brain regional cortical thickness. Neuroimage 2010; 50(2):589-99. doi: 10.1016/j.neuroimage.2009.12.047. doi: 10.1016/j.neuroimage.2009.12.047 pmid: 20026220
20.	Qureshi AI, Janardhan V, Bennett SE, et al. Who should be screened for asymptomatic carotid artery stenosis? Experience from the Western New York Stroke Screening Program. J Neuroimaging 2001; 11(2):105-11. doi: 10.1111/j.1552-6569.2001.tb00019.x. doi: 10.1111/j.1552-6569.2001.tb00019.x pmid: 11296578
21.	Jauk S, Kramer D, Quehenberger F, et al. Information adapted machine learning models for prediction in clinical workflow. Stud Health Technol Inform 2019; 260:65-72. doi: 10.3233/978-1-61499-971-3-65. pmid: 31118320
22.	Weng SF, Reps J, Kai J, et al. Can machine-learning improve cardiovascular risk prediction using routine clinical data? PLoS One 2017; 12(4):e0174944. doi: 10.1371/journal.pone.0174944. doi: 10.1371/journal.pone.0174944 pmid: 28376093
23.	Kourou K, Exarchos TP, Exarchos KP, et al. Machine learning applications in cancer prognosis and prediction. Comput Struct Biotechnol J 2014; 13:8-17. doi: 10.1016/j.csbj.2014.11.005. doi: 10.1016/j.csbj.2014.11.005 pmid: 25750696
24.	Hu X, Reaven PD, Saremi A, et al. Machine learning to predict rapid progression of carotid atherosclerosis in patients with impaired glucose tolerance. EURASIP J Bioinform Syst Biol 2016; 2016(1):14. doi: 10.1186/s13637-016-0049-6. doi: 10.1186/s13637-016-0049-6 pmid: 27642290
25.	Forsblad J, Gottsater A, Matzsch T, et al. Predictors of carotid endarterectomy in middle-aged individuals. Vasc Med 2001; 6(2):81-5. doi: 10.1177/1358836X0100600203. pmid: 11530969
26.	Zmysłowski A, Szterk A. Current knowledge on the mechanism of atherosclerosis and pro-atherosclerotic properties of oxysterols. Lipids Health Dis 2017; 16(1):188. doi: 10.1186/s12944-017-0579-2. doi: 10.1186/s12944-017-0579-2 pmid: 28969682
27.	Wiegman A, Gidding SS, Watts GF, et al. Familial hypercholesterolaemia in children and adolescents: gaining decades of life by optimizing detection and treatment. Eur Heart J 2015; 36(36):2425-37. doi: 10.1093/eurheartj/ehv157. doi: 10.1093/eurheartj/ehv157 pmid: 26009596
28.	Defesche JC, Gidding SS, Harada-Shiba M, et al. Familial hypercholesterolaemia. Nat Rev Dis Primers 2017; 3:17093. doi: 10.1038/nrdp.2017.93. pmid: 29219151
29.	Yin J, Yu C, Liu H, et al. A model to predict unstable carotid plaques in population with high risk of stroke. BMC Cardiovasc Disord 2020; 20(1):164. doi: 10.1186/s12872-020-01450-z. doi: 10.1186/s12872-020-01450-z pmid: 32264828

Characteristics	ACS (n=326)	Non ACS (n=2515)	Sub-total	χ² or t	P
Sex-male [n(%)]	175 (53.7)	1150 (45.7)	1325 (46.6)	7.339	<0.01
Age (years, mean±SD)	64.0±7.7	60.1±8.7	60.6±8.6	7.637	<0.001
Education-primary school [n(%)]	132 (40.5)	1000 (39.8)	1132 (39.8)	0.064	0.800
Marriage-married [n(%)]	292 (89.6)	1737 (69.1)	2029 (71.4)	59.444	<0.001
Physical exam index (mean±SD)
BMI	25.0±3.3	26.0±3.2	25.7±3.2	-4.169	<0.001
Height (cm)	164.3±7.5	163.2±28.0	163.3±26.5	0.712	0.476
Waistline (cm)	82.6±17.2	67.1±36.0	68.9±34.6	7.664	<0.001
Weight (kg)	67.4±9.9	68.1±9.3	68.0±9.4	-1.320	0.187
Concurrent diseases [n(%)]
Atrial fibrillation	38 (11.7)	149 (5.9)	187 (6.6)	15.420	<0.001
DM	107 (32.8)	601 (23.9)	708 (24.9)	12.288	0.001
Hypertension	275 (84.4)	1971 (78.4)	2246 (79.1)	6.246	0.05
Hypercholesterolemia	129 (39.6)	961 (38.2)	1090 (38.4)	0.226	0.635
Life style [n(%)]
Overweight	84 (25.8)	1171 (46.6)	1255 (44.2)	50.599	<0.001
Current smoking	118 (36.2)	855 (34.0)	973 (34.2)	0.620	0.431
Current drinking	77 (23.6)	312 (12.4)	389 (13.7)	30.710	<0.001
Past smoking	62 (19.0)	520 (20.7)	582 (20.5)	0.487	0.485
Diet-positive	47 (14.4)	449 (17.9)	496 (17.5)	2.364	0.124
Lack Physical activity	235 (72.1)	1935 (76.9)	2170 (76.4)	3.767	0.053
Family history [n(%)]
Stroke	86 (26.4)	798 (31.7)	884 (31.1)	3.853	<0.05
CHD	31 (9.5)	119 (4.7)	150 (5.3)	13.172	<0.001
DM	48 (14.7)	241 (9.6)	289 (10.2)	8.349	<0.01
Dyslipidemia	2 (0.6)	71 (2.8)	73 (2.6)	4.780	<0.05
Laboratory results (mmol/L, mean±SD)
FPG	6.6±2.1	6.1±1.8	6.1±1.9	4.503	<0.001
Hcy	11.3±7.7	11.1±7.7	11.1±7.7	0.506	0.613
HDL-c	1.4±0.4	1.6±0.8	1.5±0.8	-4.039	<0.001
LDL-c	3.2±0.8	3.0±0.8	3.0±0.8	4.878	<0.001
TC	5.1±0.9	5.1±1.0	5.1±1.0	1.006	0.315
TG	1.8±1.3	1.7±1.3	1.7±1.3	0.327	0.744

Rank	Risk factor	Weight	Rank	Risk factor	Weight
1	Fh-Dyslipidemia	0.118606411	16	Dyslipidemia	0.014717
2	LDL-c	0.076347664	17	Education level	0.013718
3	HDL-c	0.071114458	18	Marriage	0.013375
4	Age	0.061408055	19	Fh-DM	0.01327
5	BMI	0.061151368	20	Current drinking	0.013036
6	Waistline	0.059055337	21	Fh-CHD	0.011855
7	Glucose	0.05701681	22	DM	0.01076
8	Hcy	0.054882752	23	Fh-Stroke	0.007797
9	TC	0.054851202	24	Current smoking	0.007762
10	Height	0.049662527	25	Lack physical activity	0.007582
11	TG	0.047633703	26	Hypertension	0.007054
12	Weight	0.047228036	27	Past smoking	0.007031
13	Occupation	0.039596269	28	Overweight	0.006679
14	PayStyle	0.035715073	29	Sex	0.005956
15	Diet	0.020405623	30	Atrial fibrillation	0.004731

[1]	陈旭, 霍晓菲, 吴哲, 陆菁菁. 人工智能在卵巢癌医学影像中的应用进展[J]. Chinese Medical Sciences Journal, 2021, 36(3): 196-203.
[2]	李佳铮, 唐磊. 影像组学在抗肿瘤药物临床试验疗效评估中的应用和挑战[J]. Chinese Medical Sciences Journal, 2021, 36(3): 187-195.
[3]	史颖欢,王乾. 人工智能赋能医学影像的现状与前景[J]. Chinese Medical Sciences Journal, 2019, 34(2): 71-75.
[4]	关健. 健康和医学领域的人工智能：期许、伦理挑战和治理[J]. Chinese Medical Sciences Journal, 2019, 34(2): 76-83.
[5]	李涛, 杨立, 张卫国, 罗春才, 黄自立, 李金锋, 李欣. 6 4层多排CT对冠状动脉旁路移植术后的中期随访：影响桥血管通畅性的危险因素研究[J]. Chinese Medical Sciences Journal, 2018, 33(2): 69-76.

优先出版

当期目次

过刊浏览

虚拟专题

作者中心

评阅者中心