Estimating individualized effectiveness of receiving successful recanalization for ischemic stroke cases using
Vahid Farmani1, Helge Kniep2, Mate E Maros3
1Galway Medical Technology Centre, Department of Mechanical and Industrial Engineering, Atlantic Technological University, Galway, Ireland.
Summary
Machine learning models predict the individual benefit of mechanical thrombectomy for ischemic stroke patients. Older age and higher stroke severity scores indicate less benefit from successful recanalization.
Area of Science:
- Neurology
- Biomedical Engineering
- Data Science
Background:
- Mechanical thrombectomy (MT) is a crucial treatment for ischemic stroke.
- Quantifying the individual causal effect of successful recanalization is challenging.
- Predictive modeling can help personalize treatment decisions.
Purpose of the Study:
- To develop machine learning models to estimate the individual effect of successful recanalization in ischemic stroke patients.
- To identify patient characteristics associated with not benefiting from successful recanalization.
Main Methods:
- Utilized random forest (RF) models to predict poor functional outcomes.
- Trained models on 1718 non-reperfusion (TICI ≤ 2a) and 10339 reperfusion (TICI ≥ 2b) patient groups.
- Calculated individual treatment effect as the difference in predicted probabilities between successful and unsuccessful recanalization scenarios.
Main Results:
- RF models demonstrated strong calibration for both groups.
- Successful recanalization reduced risk by 22.0% in the reperfused group and 19.8% in the non-reperfusion group.
- Older age, higher pre-stroke modified Rankin Scale (mRS) scores, and higher National Institutes of Health Stroke Scale (NIHSS) scores were linked to not benefiting.
Conclusions:
- Machine learning effectively estimates the individual impact of MT on ischemic stroke outcomes.
- Predictive analytics can personalize MT treatment strategies.
- Identifying non-benefit predictors aids in optimizing patient selection for MT.
Keywords:
Acute ischemic strokeIndividual causal effectMachine learningMechanical thrombectomySuccessful recanalization
