Related Experiment Video
Updated: Jul 5, 2025
14:08
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
42.6K
Mortality Prediction of Patients with Subarachnoid Hemorrhage Using a Deep Learning Model Based on an Initial Brain
Sergio García-García1, Santiago Cepeda1, Dominik Müller2
1Neurosurgery Department, Rio Hortega University Hospital, 47012 Valladolid, Spain.
Brain Sciences
|January 22, 2024
Summary
Convolutional neural networks (CNNs) can predict mortality in subarachnoid hemorrhage (SAH) patients using CT scans. This AI approach offers high accuracy, aiding clinical decision-making for SAH outcomes.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Neurology
Background:
- Subarachnoid hemorrhage (SAH) is associated with significant morbidity and mortality.
- Convolutional neural networks (CNNs) demonstrate potential for accurate predictions from medical imaging data.
Purpose of the Study:
- To develop and evaluate a CNN-based algorithm for predicting mortality in SAH patients.
- Utilize initial CT scans as the sole input for mortality prediction.
Main Methods:
- Retrospective multicentric study of SAH patients.
- CT scan images were processed using the AUCMEDI framework with a DenseNet121 architecture and transfer learning.
- The model was trained to predict mortality within the first three months post-SAH.
Main Results:
- The study analyzed 219 SAH patients (28.5% mortality rate).
- The CNN model achieved 74% accuracy, 75% F1 score, and 82% AUC in predicting mortality using only baseline CT scans.
- The model demonstrated good predictive performance solely based on initial imaging.
Conclusions:
- AI and CNNs can accurately predict SAH patient mortality using CT scans.
- Further optimization with larger datasets may enhance model performance beyond conventional clinical knowledge.
