Interpretable Machine Learning Modeling for Ischemic Stroke Outcome Prediction

Abstract

Background and PurposeMechanical thrombectomy greatly improves stroke outcomes. Nonetheless, some patients fall short of full recovery despite good reperfusion. The purpose of this study was to develop machine learning (ML) models for the pre-interventional prediction of functional outcome at 3 months of thrombectomy in acute ischemic stroke (AIS), using clinical and auto-extractable radiological information consistently available upon first emergency evaluation.Materials and MethodsA two-center retrospective cohort of 293 patients with AIS who underwent thrombectomy was analyzed. ML models were developed to predict dichotomized modified Rankin score at 90 days (mRS-90) using clinical and imaging features, both separately and combined. Conventional and experimental imaging biomarkers were quantified using automated image-processing software from non-contract computed tomography (CT) and computed tomography angiography (CTA). Shapley Additive Explanation (SHAP) was applied for model interpretability and predictor importance analysis of the optimal model.ResultsMerging clinical and imaging features returned the best results for mRS-90 prediction. The best performing classifier was Extreme Gradient Boosting (XGB) with an area under the receiver operating characteristic curve (AUC) = 84% using selected features. The most important classifying features were age, baseline National Institutes of Health Stroke Scale (NIHSS), occlusion side, degree of brain atrophy [primarily represented by cortical cerebrospinal fluid (CSF) volume and lateral ventricle volume], early ischemic core [primarily represented by e-Alberta Stroke Program Early CT Score (ASPECTS)], and collateral circulation deficit volume on CTA.ConclusionMachine learning that is applied to quantifiable image features from CT and CTA alongside basic clinical characteristics constitutes a promising automated method in the pre-interventional prediction of stroke prognosis. Interpretable models allow for exploring which initial features contribute the most to post-thrombectomy outcome prediction overall and for each individual patient outcome.

Automated quantification of atrophy and acute ischemic volume for outcome prediction in endovascular thrombectomy

“…Similarly, automated atrophy quantification strongly predicts mRS, with an odds ratio of 0.65 per 5% increase in intracranial cerebrospinal fluid volume ( 15 ). A previous machine learning study of NCCT and CTA features using e-Stroke found that age, baseline NIHSS, occlusion side, atrophy and e-ASPECTS were the best predictors for mRS at 90 days, in keeping with the results shown here ( 16 ). The clinical variables used in this study—age and presenting NIHSS—have previously been used as a prognostic score, which was also an independent predictor of outcome in EVT ( 17 ).…”

Interpretable Machine Learning Model Predicting Early Neurological Deterioration in Ischemic Stroke Patients Treated with Mechanical Thrombectomy: A Retrospective Study

“…Machine learning (ML) has emerged as a promising predictive tool in medicine and has been applied in many medical fields, such as ischemic stroke outcome prediction [ 19 , 20 , 21 , 22 ], biomedical research [ 23 ] and rehabilitation for chronic stroke survivors [ 24 ], and so on. Because in the process of modeling, machine learning can fit the complex relationship in multi-dimensional data, extract subtle information, and automatically summarize and generalize to obtain new knowledge.…”

“…Machine learning has been limited in the medical field due to its black-box nature [ 25 , 26 ]. However, the latest machine learning models have been made interpretable by Shapley Additive Explanations (SHAP), which prompts the application of ML to be further developed [ 19 , 20 ]. The SHAP is a novel, cutting-edge machine learning algorithm which can visualize the relationship between each feature and the related predictive ability and can more intuitively understand the importance of features and enhance clinical interpretability.…”

Machine Learning-based prediction of Early Neurological Deterioration after Thrombolysis in Acute Ischemic Stroke

“…In recent years, machine learning (ML) has gradually been widely used in the medical field, and several prognostic and diagnostic models have been developed for ischemic stroke patients. 12,13 Optimization of models by fitting machine learning to big data has shown good predictive performance. Such ML based modeling solutions can help clinicians to better individualize their treatments.…”