Development and internal validation of a machine learning model for predicting intracranial infection after spontaneous intracerebral hemorrhage: a two-center retrospective study
Retrospective study develops and tests a machine learning model predicting intracranial infection in patients post-ICH, indicating potential clinical utility.
Key Points
The study aims to develop and validate a machine learning model for early prediction of intracranial infection following spontaneous intracerebral hemorrhage.
Two-center retrospective design including 1,317 patients with spontaneous ICH from 2015 to 2024.
Used baseline clinical, laboratory, and radiological variables within 24 hours of admission to construct the model.
Compared twelve machine learning algorithms, identifying a Light Gradient Boosting Machine as the best performer.
Intracranial infection occurred in 165 patients (12.5%).
The LGBM model demonstrated excellent discrimination with an AUC of 0.923.
Model-predicted risk showed an independent association with higher ICI risk and unfavorable functional outcomes at 180 days.