First-pass reperfusion after endovascular thrombectomy: a real-world analysis with explainable machine learning for...

Area of Science:

• Neurology
• Medical Imaging
• Machine Learning in Healthcare

Background:

• Endovascular thrombectomy is a critical treatment for acute ischemic stroke.
• Optimizing first-pass reperfusion (FPE) is crucial for patient outcomes.
• Predictive tools for intra-procedural decision support are needed.

Purpose of the Study:

• To identify determinants of FPE in endovascular thrombectomy.
• To develop an explainable machine learning (ML) framework for predicting FPE.
• To provide intra-procedural decision support for clinicians.

Main Methods:

• Retrospective analysis of 204 acute ischemic stroke patients treated with endovascular thrombectomy.
• Development and validation of six ML models using clinical, imaging, and procedural variables.
• Application of SHAP and partial dependence analyses for model interpretability.

Main Results:

• First-pass reperfusion (FPE) was achieved in 46.08% of patients.
• Favorable collateral status, lower systolic blood pressure, and higher ASPECTS scores were associated with FPE.
• Explainable ML identified collateral circulation, systolic blood pressure, hypertension, and age as key predictors.

Conclusions:

• An internally validated, explainable ML framework can estimate FPE likelihood using routine variables.
• Findings are preliminary due to limited events and lack of external validation.
• Further multicenter validation is required before clinical implementation.