Development and External Validation of a Deep Learning Model to Predict Mortality in Aneurysmal Subarachnoid Hemorrhage Using Admission Computed Tomography

Key Points

• This research aims to develop a deep learning model to predict 90-day mortality in patients with aneurysmal subarachnoid hemorrhage using admission computed tomography scans.
• Development of a fully automated deep learning model based on admission non-contrast computed tomography (NCCT) scans.
• Validation of the model's predictive accuracy against conventional clinical models.
• Objective assessment without additional data collection beyond standard imaging.
• The deep learning model demonstrates comparable discrimination and calibration for predicting mortality as established clinical models.
• The model provides objective, reproducible risk stratification at the point of care.
• It serves as a complementary decision-support tool in clinical settings.