Automatic Diagnosis Based on Spatial Information Fusion Feature for Intracranial Aneurysm

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Radiology

Background:

• Digital Subtraction Angiography (DSA) is the gold standard for intracranial aneurysm diagnosis, but manual interpretation carries a risk of misdiagnosis.
• Computer-aided diagnosis (CAD) systems can assist radiologists, potentially improving accuracy and efficiency in detecting intracranial aneurysms.

Purpose of the Study:

• To develop and evaluate a deep learning-based computer-aided diagnosis (CAD) method for detecting intracranial aneurysms using 3D Rotational Angiography (3D-RA) 2D image sequences.
• To leverage spatial information fusion (SIF) to enhance feature extraction from 2D image sequences, avoiding the computational cost of 3D Convolutional Neural Networks (CNNs).

Main Methods:

• A novel spatial information fusion (SIF) method was developed to embed temporal and spatial contextual information from consecutive 3D-RA frames into a single 2D image.
• The fused 2D images were trained using a 2D Convolutional Neural Network (CNN) for intracranial aneurysm detection.
• Morphological differences between consecutive frames were utilized as the primary feature for detection, addressing challenges posed by similar features between aneurysms and vascular structures.

Main Results:

• The proposed deep learning method achieved a high diagnostic accuracy of 98.89% for intracranial aneurysm detection.
• The system demonstrated excellent sensitivity (99.38%) and specificity (98.19%).
• The SIF method proved feasible and effective for analyzing 2D image sequences, offering a computationally efficient alternative to 3D CNNs.

Conclusions:

• The developed deep learning method utilizing spatial information fusion (SIF) is a feasible and highly accurate approach for the auxiliary diagnosis of intracranial aneurysms from 3D-RA 2D image sequences.
• This CAD system has the potential to significantly aid radiologists, leading to quicker treatment decisions, improved patient outcomes, and reduced healthcare costs.
• The focus on inter-frame morphological differences offers a robust strategy for distinguishing aneurysms from overlapping vascular structures in medical imaging.