AUTHOR=Jabal Mohamed Sobhi , Joly Olivier , Kallmes David , Harston George , Rabinstein Alejandro , Huynh Thien , Brinjikji Waleed 

TITLE=Interpretable Machine Learning Modeling for Ischemic Stroke Outcome Prediction

JOURNAL=Frontiers in Neurology

VOLUME=Volume 13 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/neurology/articles/10.3389/fneur.2022.884693

DOI=10.3389/fneur.2022.884693

ISSN=1664-2295

ABSTRACT=Background and Purpose: Mechanical thrombectomy greatly improves stroke outcomes. Yet, some patients fall short of full recovery despite good reperfusion. The study purpose was to develop machine learning models for pre-interventional prediction of functional outcome at 3 months of thrombectomy in acute ischemic stroke, using clinical and auto-extractable radiological information consistently available upon first emergency evaluation.  
Materials and Methods:  A two-center retrospective cohort of 293 patients with acute ischemic stroke who underwent thrombectomy was analyzed. Machine learning models were developed to predict dichotomized modified Rankin score at 90 days using clinical and imaging features, both separately and combined. Conventional and experimental imaging biomarkers were quantified using automated image processing software from non-contract CT and CTA. Shapley additive explanation was applied for model interpretability and predictor importance analysis of the optimal model.
Results: Merging clinical and imaging features returned the best results for mRS-90 prediction. The best performing classifier was Extreme Gradient Boosting with an AUC = 84% using selected features. The most important classifying features were age, baseline NIHSS, occlusion site, degree of brain atrophy (primarily represented by cortical CSF volume and lateral ventricle volume), early ischemic core (primarily represented by e-ASPECTS), and collateral circulation deficit volume on CTA.
Conclusion: Machine learning applied to quantifiable image features from CT and CTA alongside basic clinical characteristics constitutes a promising automated method in pre-interventional prediction of stroke prognosis.  Interpretable models allow for exploring which initial features contribute the most to thrombectomy outcome prediction overall and for each individual patient outcome.