Deep learning for automated cerebral aneurysm detection on computed tomography images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Cerebrovascular aneurysms incidence is rising, necessitating advanced detection tools.
• Current detection methods require improvement in accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate an automated aneurysm detection model using deep learning.
• To enhance aneurysm visibility and detection performance through a novel image processing method.

Main Methods:

• Utilized a faster RCNN deep learning model with 2D nearby projection (NP) images derived from 3D CT angiography (CTA).
• Developed a novel NP method to enhance aneurysm visualization.
• Trained and tested the model on data from 311 patients (352 aneurysms) across three medical centers.

Main Results:

• Achieved an overall sensitivity of 91.8% for aneurysm detection.
• Demonstrated high sensitivity (96.7%) for aneurysms larger than 3 mm, reaching state-of-the-art performance.
• Detection time was under 25 seconds per case, with consistent performance across different aneurysm locations.

Conclusions:

• Successfully developed a deep learning model for automated aneurysm detection.
• The model exhibits robust performance across various aneurysm sizes and locations.
• This automated detection tool has the potential to significantly enhance clinical performance in diagnosing aneurysms.