CRANet: a comprehensive residual attention network for intracranial aneurysm image classification

Area of Science:

• Medical Imaging and Artificial Intelligence
• Neurology and Neurosurgery
• Radiology

Background:

• Subarachnoid hemorrhage (SAH) due to intracranial aneurysm rupture has high mortality.
• Magnetic Resonance Imaging (MRI) is vital for detecting and characterizing intracranial aneurysms.
• Increasing workload of analyzing aneurysm images can lead to diagnostic errors.

Purpose of the Study:

• To develop a simple and effective deep learning model for accurate intracranial aneurysm detection.
• To improve diagnostic accuracy and reduce the workload for radiologists and clinicians.
• To enhance the early detection of aneurysms, potentially improving patient outcomes.

Main Methods:

• Proposed a Comprehensive Residual Attention Network (CRANet) for aneurysm detection.
• Utilized a residual network architecture to extract key aneurysm features from MRI scans.
• Evaluated the model's performance on a dedicated test dataset.

Main Results:

• The CRANet model demonstrated effective detection of intracranial aneurysms.
• Achieved a high accuracy rate of 97.81% on the test set.
• Reached a recall rate of 94%, significantly improving aneurysm detection.

Conclusions:

• The proposed CRANet model offers a promising solution for automated aneurysm detection.
• CRANet effectively enhances the accuracy and efficiency of aneurysm diagnosis from MRI.
• This AI-driven approach has the potential to improve SAH management and reduce mortality.