A Machine Learning Model Based on Brain Imaging and Clinical Features for Predicting Atrial Fibrillation Detected after Stroke

doi:10.3969/j.issn.1673-5765.2025.04.002

Abstract

Abstract: Objective To investigate the predictive value of a machine learning model based on brain imaging and clinical features in patients with atrial fibrillation detected after stroke.
Methods A retrospective cohort design was used in this study. Data was derived from the ischemic stroke and TIA patients enrolled in the China national stroke registry Ⅲ from August 2015 to March 2018. Patients were divided into two groups according to the systematic collection of past medical history records, electrocardiogram, and 24-hour Holter monitoring results during hospitalization: the sinus rhythm group and the atrial fibrillation detected after stroke group. Firstly, a pre-trained nnUNet deep learning framework was applied for standardized preprocessing and automated lesion segmentation of DWI data. Subsequently, 960 quantitative imaging features encompassing eight categories, including morphological characteristics, first-order statistics, and advanced texture features, were extracted using the PyRadiomics open-source package. During the feature engineering stage, the Spearman’s rank correlation coefficient analysis was applied (preset threshold |ρ|>0.8) to eliminate highly collinear features. After retaining independent features, the least absolute shrinkage and selection operator (LASSO) regression algorithm was used for feature selection and to construct a joint prediction model. The model performance was internally validated via five-fold cross-validation, and the AUC of the ROC curve was used as the primary evaluation indicator. Finally, the SHapley Additive exPlanations framework was used to analyze the importance of features.
Results A total of 1464 ischemic stroke patients were included, with an average age of (64.5±11.1) years, including 498 patients with atrial fibrillation detected after stroke and 966 patients with sinus rhythm. The average AUC of five-fold cross-validation of the prediction model for atrial fibrillation detected after stroke constructed using 15 clinical features was 0.71 (95%CI 0.67-0.74). Clinical and imaging features were fused to form 975 multimodal features, with an average AUC of 0.73 (95%CI 0.70-0.76). Using the LASSO algorithm for feature selection, 31 multimodal features (including 25 imaging and 6 clinical features) were obtained after screening, with an average AUC of 0.73 (95%CI 0.70-0.77).
Conclusions The machine learning model based on brain imaging and clinical features can effectively predict atrial fibrillation detected after stroke, and can be further applied in clinical practice.

Key words: Ischemic stroke; Atrial fibrillation; Machine learning

摘要： 目的探讨基于患者脑影像及临床特征的机器学习模型对缺血性卒中后心房颤动的预测价值。
方法本研究采用回顾性队列研究设计，数据源于中国国家卒中登记Ⅲ，纳入2015年8月—2018年3月收录的缺血性卒中和TIA病例。通过系统收集的既往病史记录，住院期间的心电图检查及24 h动态心电监测结果将患者分为窦性心律组和缺血性卒中后心房颤动组。首先应用预训练的nnUNet深度学习框架对DWI数据进行标准化预处理及病灶自动分割。随后基于PyRadiomics开源软件包提取病灶的影像组学特征，涵盖形态学特征、一阶统计量及高阶纹理特征等8类共960项定量影像特征。特征工程阶段，首先应用斯皮尔曼等级相关系数分析（预设阈值|ρ|>0.8）剔除高度共线性特征，保留独立特征后，采用最小绝对收缩和选择算子（least absolute shrinkage and selection operator，LASSO）回归算法进行特征选择并构建联合预测模型。模型性能通过5折交叉验证进行内部验证，采用ROC曲线的AUC作为主要评价指标。最终通过SHAP（SHapley Additive exPlanations）框架解析特征重要性。
结果共纳入1464例缺血性卒中患者，平均年龄（64.5±11.1）岁，缺血性卒中后心房颤动患者498例，窦性心律患者966例。基于15项临床特征构建的缺血性卒中后心房颤动预测模型的5折交叉验证平均AUC值为0.71（95%CI 0.67～0.74）；将临床特征和影像组学特征融合，构成975项多模态特征，平均AUC值为0.73（95%CI 0.70～0.76）；使用LASSO方法进行特征筛选，筛选后得到31项多模态特征（包括25项影像组学特征和6项临床特征），平均AUC值为0.73（95%CI 0.70～0.77）。
结论基于脑影像及临床特征的机器学习模型能有效预测缺血性卒中后心房颤动，可进一步应用于临床工作中。

关键词: 缺血性卒中; 心房颤动; 机器学习

CLC Number:

R74
R318

ZHANG Liyuan, LIU Tao, JIANG Yong, LI Zixiao, WANG Yongjun, YANG Xiaomeng. A Machine Learning Model Based on Brain Imaging and Clinical Features for Predicting Atrial Fibrillation Detected after Stroke[J]. Chinese Journal of Stroke, 2025, 20(4): 392-400.

张栗源, 刘涛, 姜勇, 李子孝, 王拥军, 杨晓萌. 基于脑影像及临床特征的机器学习模型预测缺血性卒中后心房颤动[J]. 中国卒中杂志, 2025, 20(4): 392-400.

References

[1] CERASUOLO J O，CIPRIANO L E，SPOSATO L A. The complexity of atrial fibrillation newly diagnosed after ischemic stroke and transient ischemic attack：advances and uncertainties[J]. Curr Opin Neurol，2017，30（1）：28-37.
[2] SPOSATO L A，CHATURVEDI S，HSIEH C Y，et al. Atrial fibrillation detected after stroke and transient ischemic attack：a novel clinical concept challenging current views[J/OL]. Stroke，2022，53（3）：e94-e103[2023-06-10]. https://doi.org/10.1161/STROKEAHA.121.034777.
[3] TSIVGOULIS G，PALAIODIMOU L，TRIANTAFYLLOU S，et al. Prolonged cardiac monitoring for stroke prevention：a systematic review and meta-analysis of randomized-controlled clinical trials[J]. Eur Stroke J，2023，8（1）：106-116.
[4] KISHORE A K，HOSSAIN M J，CAMERON A，et al. Use of risk scores for predicting new atrial fibrillation after ischemic stroke or transient ischemic attack—a systematic review[J]. Int J Stroke，2022，17（6）：608-617.
[5] TING D S W，LIU Y，BURLINA P，et al. AI for medical imaging goes deep[J]. Nat Med，2018，24（5）：539-540.
[6] HARMON D M，SEHRAWAT O，MAANJA M，et al. Artificial intelligence for the detection and treatment of atrial fibrillation[J/OL]. Arrhythm Electrophysiol Rev，2023，12：e12[2023-06-10]. https://doi.org/10.15420/aer.2022.31.
[7] CHAMBERLAIN A M，ROGER V L，NOSEWORTHY P A，et al. Identification of incident atrial fibrillation from electronic medical records[J/OL]. J Am Heart Assoc，2022，11（7）：e023237[2023-06-10]. https://doi.org/10.1161/JAHA.121.023237.
[8] WANG Y J，JING J，MENG X，et al. The third China national stroke registry（CNSR-Ⅲ）for patients with acute ischaemic stroke or transient ischaemic attack：design，rationale and baseline patient characteristics[J]. Stroke Vasc Neurol，2019，4（3）：158-164.
[9] VAN GRIETHUYSEN J J M，FEDOROV A，PARMAR C，et al. Computational radiomics system to decode the radiographic phenotype[J/OL]. Cancer Res，2017，77（21）：e104-e107[2023-06-10]. https://doi.org/10.1158/0008-5472.CAN-17-0339.
[10] MUTHUKRISHNAN R，ROHINI R. LASSO：a feature selection technique in predictive modeling for machine learning[C]//2016 IEEE International Conference on Advances in Computer Applications（ICACA）. Coimbatore，India：IEEE，2016：18-20.
[11] PROKHORENKOVA L，GUSEV G，VOROBEV A，et al. CatBoost：unbiased boosting with categorical features[EB/OL].（2019-01-20）[2023-06-05]. https://arxiv.org/abs/1706.09516.
[12] LUNDBERG S M，LEE S I. A unified approach to interpreting model predictions[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook，New York，USA：Curran Associates Inc. ，2017：4768-4777.
[13] UPHAUS T，WEBER-KRÜGER M，GROND M，et al. Development and validation of a score to detect paroxysmal atrial fibrillation after stroke[J/OL]. Neurology，2019，92（2）：e115-e124[2023-06-10]. https://doi.org/10.1212/WNL.0000000000006727.
[14] SHAROBEAM A，CHURILOV L，PARSONS M，et al. Patterns of infarction on MRI in patients with acute ischemic stroke and cardio-embolism：a systematic review and meta-analysis[J/OL]. Front Neurol，2020，11：606521[2023-06-10]. https://doi.org/10.3389/fneur.2020.606521.
[15] MAYERHOEFER M E，MATERKA A，LANGS G，et al. Introduction to radiomics[J]. J Nucl Med，2020，61（4）：488-495.
[16] CHEN Y，HE Y W，JIANG Z Y，et al. Ischemic stroke subtyping method combining convolutional neural network and radiomics[J]. J Xray Sci Technol，2023，31（2）：223-235.
[17] CHEN Q，XIA T Y，ZHANG M Y，et al. Radiomics in stroke neuroimaging：techniques，applications，and challenges[J]. Aging Dis，2021，12（1）：143-154.
[18] WU O，WINZECK S，GIESE A K，et al. Big data approaches to phenotyping acute ischemic stroke using automated lesion segmentation of multi-center magnetic resonance imaging data[J]. Stroke，2019，50（7）：1734-1741.