Carlos Fernandez-Lozano, Pablo Hervella, Virginia Mato‐Abad, Manuel Rodríguez‐Yáñez, Sonia Suárez, Iria López‐Dequidt, Ana Estany‐Gestal, Tomás Sobrino et al.
importance 7/10ai assistedAcute Ischemic Stroke Managementstroke medicinestage: research
Problem
Clinicians lack tools to predict which stroke patients will die or suffer morbidity three months after admission.
Why it matters
Accurate early prediction of stroke death and disability could guide treatment for thousands of patients.
What the AI did
A Random Forest model trained on clinical, biochemical, and neuroimaging variables from 6022 prospectively registered patients predicts 3-month mortality and morbidity.
Results
Model built on 6022 patients (4922 ischemic stroke, 1100 ICH); IS+ICH group most stable for mortality prediction at 0.904 (metric truncated in abstract).
Impact
Identified NIHSS at 24/48h and admission temperature as key predictors, enabling outcome estimation by stroke subtype.
Record
Venue
Scientific Reports
Published
2021
Work type
article
Authors
Carlos Fernandez-Lozano, Pablo Hervella, Virginia Mato‐Abad, Manuel Rodríguez‐Yáñez, Sonia Suárez, Iria López‐Dequidt, Ana Estany‐Gestal, Tomás Sobrino et al.
stroke, random forest, mortality prediction, machine learning, neuroimaging
Importance rating
7/ 10
significant — thousands to millions affected
Stroke outcome prediction could guide treatment for thousands of patients; model built on 6022 prospectively registered patients with AUC up to ~0.90 for mortality.
Basis: stated.
The source itself states the scale.
Stated scale: 6022 prospectively registered stroke patients. Method and limits.