Abstract:
Objective: This study aims to develop and validate the Futile Recanalization Prediction Score (FRPS), a novel tool designed to predict the severity risk of FR and aid in pre- and post-EVT risk assessments. Methods: The FRPS was developed using a rigorous process involving the selection of predictor variables based on clinical relevance and potential impact. Initial equations were derived from previous meta-analyses and refined using various statistical techniques. We employed machine learning algorithms, specifically random forest regression, to capture nonlinear relationships and enhance model performance. Cross-validation with five folds was used to assess generalizability and model fit. Results: The final FRPS model included variables such as age, sex, atrial fibrillation (AF), hypertension (HTN), diabetes mellitus (DM), hyperlipidemia, cognitive impairment, pre-stroke modified Rankin Scale (mRS), systolic blood pressure (SBP), onset-to-puncture time, sICH, and NIHSS score. The random forest model achieved a mean R-squared value of approximately 0.992. Severity ranges for FRPS scores were defined as mild (FRPS < 66), moderate (FRPS 66-80), and severe (FRPS > 80). Conclusions: The FRPS provides valuable insights for treatment planning and patient management by predicting the severity risk of FR. This tool may improve the identification of candidates most likely to benefit from EVT and enhance prognostic accuracy post-EVT. Further clinical validation in diverse settings is warranted to assess its effectiveness and reliability.
摘要:
目的:本研究旨在开发和验证徒劳再化预测评分(FRPS),一种新的工具,旨在预测FR的严重风险,并有助于EVT前后的风险评估。方法:使用严格的过程开发FRPS,该过程涉及根据临床相关性和潜在影响选择预测变量。初始方程来自先前的荟萃分析,并使用各种统计技术进行了完善。我们采用了机器学习算法,特别是随机森林回归,捕获非线性关系并增强模型性能。使用五个折叠的交叉验证来评估泛化性和模型拟合。结果:最终的FRPS模型包括年龄、性别,心房颤动(AF),高血压(HTN),糖尿病(DM),高脂血症,认知障碍,卒中前改良Rankin量表(mRS),收缩压(SBP),从开始到穿刺的时间,sICH,和NIHSS得分。随机森林模型实现了约0.992的平均R平方值。FRPS评分的严重程度范围定义为轻度(FRPS<66),中等(FRPS66-80),严重(FRPS>80)。结论:FRPS通过预测FR的严重风险为治疗计划和患者管理提供了有价值的见解。该工具可以改善最有可能受益于EVT的候选者的识别,并提高EVT后的预后准确性。需要在不同环境中进行进一步的临床验证,以评估其有效性和可靠性。