A deep-learning-based model for risk assessment of ruptured intracranial aneurysms

Abstract

Intracranial aneurysm (IA) is a common cerebrovascular disease, with approximately 85% of spontaneous subarachnoid hemorrhage (SAH) cases caused by its rupture. The mortality rate exceeds 40%, and it may lead to lifelong cognitive impairment. Accurate assessment of IA rupture risk is crucial for clinical decision-making.We propose a deep learning-based MambaEPA model that integrates Mamba and Efficient Paired Attention (EPA) blocks for precise IA rupture risk assessment. Unlike existing methods, our model not only focuses on aneurysm morphology but also considers the impact of the surrounding environment on rupture risk. The Mamba module captures long-range dependencies at both the channel and spatial levels, effectively modeling the complex relationships between aneurysms and surrounding tissues. Meanwhile, the EPA block enhances feature extraction through spatial and channel attention mechanisms. Experimental results demonstrate that the proposed model outperforms existing methods in both prediction performance and computational efficiency, providing a novel solution for automated IA rupture risk assessment. This study expands the application of deep learning in medical image analysis, offering significant clinical value.

Ähnliche Arbeiten