Scholarlink

Introduction <p>Flow diverters (FD) have gradually become the preferred treatment option for complex and large intracranial aneurysms. Postoperative thromboembolic events (TEEs) are among the most common complications associated with endovascular treatment. However, widely applicable predictive tools for the occurrence of TEEs are currently lacking.</p> Methods <p>This retrospective study included clinical data from 377 patients (a total of 451 aneurysms) treated with flow diverters at two neurointerventional centers between June 2018 and September 2022. Thirty-nine baseline patient characteristics were included as clinical variables. The primary endpoint was the occurrence of postoperative ischemic events. The dataset was randomly divided into a training set (80%) and a testing set (20%). We performed fivefold cross-validation and applied Lasso regression to the training set to identify the most informative features. Multiple machine learning (ML) algorithms were employed to construct predictive models. Model performance was evaluated on the testing set using the area under the receiver operating characteristic curve (AUC-ROC), the area under the precision–recall curve (AUC-PR), and calibration plots. SHapley Additive exPlanations (SHAP) analysis was used to visualize feature contributions and to interpret individual case predictions.</p> Results <p>Among 377 patients, 21 (5.6%) experienced TEEs. A machine learning model incorporating 10 variables was developed, with the support vector machine (SVM) model demonstrating the best performance—achieving an AUC-ROC of 0.96 and an AUC-PR of 0.88 in validation. The key predictive factors included aneurysm width, low-density lipoprotein (LDL) levels, hypertension, aneurysm location, triglycerides (TG), and diabetes. Additionally, a web-based tool was developed to assist clinicians in applying the model in practice.</p> Conclusions <p>We developed a machine learning model to predict the risk of TEEs following FD implantation for intracranial aneurysms, and demonstrated its clinical potential through internal validation. This tool can assist neurointerventionalists in estimating the probability of TEE occurrence based on patient clinical data and aneurysm characteristics, enabling the development of personalized treatment strategies.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine Learning for Predicting Thromboembolic Events Following Flow Diverter Treatment of Intracranial Aneurysms: A Multicenter Retrospective Study

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月27日

查看全文
  • Yunpeng Lin,
  • Xiaoning Liu,
  • Bingcheng Ren,
  • Jiwen Wang,
  • Yang Li,
  • Xiangbo Liu,
  • Yidi Wang,
  • Fushun Xiao,
  • Shiqing Mu

摘要

Introduction

Flow diverters (FD) have gradually become the preferred treatment option for complex and large intracranial aneurysms. Postoperative thromboembolic events (TEEs) are among the most common complications associated with endovascular treatment. However, widely applicable predictive tools for the occurrence of TEEs are currently lacking.

Methods

This retrospective study included clinical data from 377 patients (a total of 451 aneurysms) treated with flow diverters at two neurointerventional centers between June 2018 and September 2022. Thirty-nine baseline patient characteristics were included as clinical variables. The primary endpoint was the occurrence of postoperative ischemic events. The dataset was randomly divided into a training set (80%) and a testing set (20%). We performed fivefold cross-validation and applied Lasso regression to the training set to identify the most informative features. Multiple machine learning (ML) algorithms were employed to construct predictive models. Model performance was evaluated on the testing set using the area under the receiver operating characteristic curve (AUC-ROC), the area under the precision–recall curve (AUC-PR), and calibration plots. SHapley Additive exPlanations (SHAP) analysis was used to visualize feature contributions and to interpret individual case predictions.

Results

Among 377 patients, 21 (5.6%) experienced TEEs. A machine learning model incorporating 10 variables was developed, with the support vector machine (SVM) model demonstrating the best performance—achieving an AUC-ROC of 0.96 and an AUC-PR of 0.88 in validation. The key predictive factors included aneurysm width, low-density lipoprotein (LDL) levels, hypertension, aneurysm location, triglycerides (TG), and diabetes. Additionally, a web-based tool was developed to assist clinicians in applying the model in practice.

Conclusions

We developed a machine learning model to predict the risk of TEEs following FD implantation for intracranial aneurysms, and demonstrated its clinical potential through internal validation. This tool can assist neurointerventionalists in estimating the probability of TEE occurrence based on patient clinical data and aneurysm characteristics, enabling the development of personalized treatment strategies.