🧠 A new benchmark for post-stroke outcome prediction is here.
Olsson et al. (2026, Scientific Reports) trained explainable ML models on 506 acute ischemic stroke patients across four stroke units in Sweden, predicting 3-month functional outcome (mRS > 2):
📊 AUROC 0.900–0.906
📈 AUPRC 0.773 ± 0.080 (MLP, best model)
🎯 Sensitivity 0.655 | Specificity 0.920
Top signals 👇
⚡ NIHSS dominated clinical features
🧬 Brain-derived tau (BD-tau) led blood biomarkers
🔍 SHAP interpretability across XGBoost + MLP + regularized logistic regression
🔗 Full paper: https://lnkd.in/gCxQsddg
#MedicalAI #StrokeCare #ExplainableAI #PrognosticAI #DigitalHealth #SaMD #MachineLearning #Biomarkers #Neurology #ClinicalAI #HealthcareInnovation #PurpleAI