Prediction of Hemorrhagic Transformation after Ischemic Stroke Using Machine Learning

Area of Science:

• Neurology
• Medical Informatics
• Artificial Intelligence

Background:

• Hemorrhagic transformation (HT) is a significant predictor of poor outcomes following acute ischemic stroke (AIS).
• Accurate prediction of HT is crucial for optimizing patient management and therapeutic strategies.

Purpose of the Study:

• To compare the predictive performance of various machine learning (ML) algorithms for HT after AIS using only structured clinical data.
• To evaluate the efficacy of artificial neural networks (ANNs) in predicting HT occurrence in AIS patients.

Main Methods:

• A dataset of 2028 AIS patients was analyzed, with HT defined by European Co-operative Acute Stroke Study-II criteria.
• Binary logistic regression, support vector machine, extreme gradient boosting, and ANN models were trained and tested.
• Model performance was assessed using the area under the receiver operating characteristic curve (AUROC), with hyperparameter optimization via cross-validation and grid search.

Main Results:

• The ANN algorithm demonstrated the highest predictive performance with an AUROC of 0.844.
• Feature scaling and resampling strategies did not significantly enhance the ANN's predictive accuracy.
• The ANN model outperformed conventional ML algorithms in predicting HT after AIS.

Conclusions:

• Artificial neural networks show superior performance in predicting HT after AIS compared to traditional ML algorithms.
• Deep learning models hold promise for predicting critical outcomes using structured data in stroke patients.