Data-Driven Prognostication in Distal Medium Vessel Occlusions Using Explainable Machine Learning

Mert Karabacak; Burak Berksu Ozkara; Tobias D Faizy; Trevor Hardigan; Jeremy J Heit; Dhairya A Lakhani; Konstantinos Margetis; J Mocco; Kambiz Nael; Max Wintermark; Vivek S Yedavalli

doi:10.3174/ajnr.A8547

Data-Driven Prognostication in Distal Medium Vessel Occlusions Using Explainable Machine Learning

AJNR Am J Neuroradiol. 2024 Oct 28:ajnr.A8547. doi: 10.3174/ajnr.A8547. Online ahead of print.

Affiliation

¹ From the Department of Neurosurgery (M.K., T.H., K.M., J.M.), and Department of Diagnostic, Molecular and Interventional Radiology (B.B.O.), Mount Sinai Health System, New York, NY, USA; Neuroendovascular Division, Department of Radiology (T.D.F.), University Medical Center Münster, Münster, Germany; Departments of Radiology and Neurosurgery (J.J.H.), Stanford University, Palo Alto, CA, USA; Department of Neuroradiology (D.A.L.), Rockefeller Neuroscience Institute, West Virginia University, Morgantown, WV, USA; Department of Radiology (K.N.), University of California San Francisco, San Francisco, CA, USA; Department of Neuroradiology (M.W.), MD Anderson Cancer Center, Houston, TX, USA; Russell H. Morgan Department of Radiology and Radiological Sciences (V.S.Y.), Johns Hopkins Medicine, Baltimore, MD, USA.

PMID: 39443146
DOI: 10.3174/ajnr.A8547

Abstract

Background and purpose: Distal medium vessel occlusions (DMVOs) are estimated to cause acute ischemic stroke (AIS) in 25-40% of cases. Prognostic models can inform patient counseling and research by enabling outcome predictions. However, models designed specifically for DMVOs are lacking.

Materials and methods: This retrospective study developed a machine learning model to predict 90-day unfavorable outcome [defined as a modified Rankin Scale (mRS) score of 3-6] in 164 primary DMVO patients. A model developed with the TabPFN algorithm utilized selected clinical, laboratory, imaging, and treatment data with the Least Absolute Shrinkage and Selection Operator feature selection. Performance was evaluated via 5-repeat 5-fold cross-validation. Model discrimination and calibration were evaluated. SHapley Additive Explanations (SHAP) identified influential features. A web application deployed the model for individualized predictions.

Results: The model achieved an area under the receiver operating characteristic curve of 0.815 (95% CI: 0.79-0.841) for predicting unfavorable outcome, demonstrating good discrimination, and a Brier score of 0.19 (95% CI: 0.177-0.202), demonstrating good calibration. SHAP analysis ranked admission National Institutes of Health Stroke Scale (NIHSS) score, premorbid mRS, type of thrombectomy, modified thrombolysis in cerebral infarction score, and history of malignancy as top predictors. The web application enables individualized prognostication.

Conclusions: Our machine learning model demonstrated good discrimination and calibration for predicting 90-day unfavorable outcomes in primary DMVO strokes. This study demonstrates the potential for personalized prognostic counseling and research to support precision medicine in stroke care and recovery.

Abbreviations: DMVO = Distal medium vessel occlusion; AIS = acute ischemic stroke; mRS = modified Rankin Scale; SHAP = SHapley Additive Explanations; NIHSS = National Institutes of Health Stroke Scale; ST = stroke thrombectomy; TRIPOD = Transparent Reporting of Multivariable Prediction Models for Individual Prognosis or Diagnosis; CT = computed tomography; CTP = CT perfusion; MRI = magnetic resonance imaging; CTA = CT angiography; DVT = deep vein thrombosis; PE = pulmonary emboli; TIA = transient ischemic attack; BMI = body mass index; ALP = alkaline phosphatase; ALT = alanine transaminase; AST = aspartate aminotransferase; NCCT-ASPECTS = Alberta Stroke Program Early CT Score; IVT = intravenous thrombolysis; mTICI = modified thrombolysis in cerebral infarction; ER = emergency room; kNN = k-nearest neighbor; LASSO = Least Absolute Shrinkage and Selection Operator; PDPs = partial dependence plots; ROC = receiver operating characteristic; PRC = precision-recall curve; AUROC = area under the ROC curve; AUPRC = area under the PRC; CI = confidence interval.