Related Experiment Video
Updated: Jun 27, 2025
14:08
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
42.6K
Evaluating Deep Learning Techniques for Detecting Aneurysmal Subarachnoid Hemorrhage: A Comparative Analysis of
Mustafa Umut Etli1, Muhammet Sinan Başarslan2, Eyüp Varol1
1Department of Neurosurgery, Ümraniye Training And Research Hospital, İstanbul, Turkey.
World Neurosurgery
|May 6, 2024
Summary
Deep learning models, particularly EfficientNetB4, accurately detect subarachnoid hemorrhage (SAH) from CT scans. This transfer learning approach offers a faster, more effective method for diagnosing SAH compared to traditional techniques.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Neurology
Background:
- Subarachnoid hemorrhage (SAH) detection is crucial for patient outcomes.
- Traditional SAH detection methods can be time-consuming.
- Machine learning offers a promising avenue for improving SAH diagnosis.
Purpose of the Study:
- To evaluate the effectiveness of convolutional neural networks (CNNs) and transfer learning models for SAH detection.
- To differentiate between aneurysmal SAH and non-aneurysmal SAH using deep learning.
- To assess the performance of novel transfer learning approaches in SAH diagnosis.
Main Methods:
- Utilized 15,600 CT images from 123 aneurysmal SAH and 80 non-aneurysmal SAH patients.
- Employed Inception-V3, EfficientNetB4, single-layer CNN, and three-layer CNN models.
- Customized transfer learning models and trained them using the Adam optimizer on Google Collaboratory.
Main Results:
- EfficientNetB4 achieved the highest accuracy (99.92%), F-score (99.82%), recall (99.92%), and precision (99.90%).
- CNN and transfer learning models showed comparable results without overfitting.
- Robust diagnostic models were developed for SAH detection.
Conclusions:
- CNN-based transfer learning models accurately diagnose SAH etiology from CT images.
- This AI approach serves as a valuable tool for clinicians, potentially reducing invasive procedures.
- Improved medical resource utilization and patient outcomes are anticipated with this technology.
