Predicting hematoma expansion using machine learning: An exploratory analysis of the ATACH 2 trial

Area of Science:

• Neurology
• Medical Imaging
• Artificial Intelligence

Background:

• Hematoma expansion (HE) is a key predictor of poor prognosis in intracerebral hemorrhage (ICH).
• Predicting HE is crucial for patient management and treatment strategies.
• Current prediction models may not fully leverage advanced computational techniques.

Purpose of the Study:

• To develop and evaluate deep learning classification models for predicting HE in ICH patients.
• To assess the performance of machine learning algorithms without relying on advanced radiological features.
• To identify key clinical variables associated with HE.

Main Methods:

• Utilized data from the Antihypertensive Treatment of Acute Cerebral Hemorrhage (ATACH-2) trial.
• Employed multiple machine learning algorithms with iterative feature selection and outcome balancing.
• Defined HE as >33% or 6 mL increase in hematoma volume within 24 hours.
• Compared models against logistic regression using AUC, accuracy, sensitivity, and specificity.

Main Results:

• The best performing model was an artificial neural network (ANN) with an AUC of 0.702.
• ANN performance was modestly better than traditional logistic regression (AUC 0.658).
• Initial hematoma volume, time to CT head, and systolic blood pressure were key predictors.

Conclusions:

• Developed machine learning models, including an ANN, to predict HE in ICH patients.
• The ANN model showed improved, though modest, predictive performance without advanced radiographic features.
• Further validation on larger, diverse datasets is necessary to generalize the models.