Scholarlink

Abstract <p>Intracranial<?tk 4?> aneurysms bring substantial health risks due to their potential to rupture, which may lead to severe morbidity and mortality. Accurate prediction of rupture risk is essential for guiding clinical decision-making and formulating effective treatment strategies. In this study, we introduce GN-Net, a novel deep learning model designed to predict the rupture risk of intracranial aneurysms by integrating geometric and neighborhood features extracted from 3D Computed Tomography Angiography (CTA) images. GN-Net comprises two distinct branches: a geometric branch that utilizes geometric deep learning to capture the local geometric structures of aneurysms and their parent arteries, and a neighborhood-aware branch that employs 3D Convolutional Neural Networks (CNNs) and Transformer encoders to model the surrounding anatomical context. We evaluated GN-Net on an internal dataset and an external dataset. This study collected 423 valid intracranial aneurysm patients from a hospital for model training and testing. Experimental results demonstrated that our method achieved the highest prediction accuracy (91.46%), and significantly outperformed the existing state-of-the-art methods. Furthermore, the proposed model was also able to improve clinicians’ diagnostic accuracy by about 7% in assessing aneurysm rupture status. The model particularly benefited less experienced clinicians, indicating its potential to bridge expertise gaps and support decision-making processes effectively. These findings underscore GN-Net’s potential as a powerful tool for improving rupture risk prediction and supporting clinicians in making informed treatment decisions. The code for this study is available at <a href="https://github.com/YouWillLikeIt/GN-Net">https://github.com/YouWillLikeIt/GN-Net</a>.<?tk 0?></p> Graphical Abstract <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

GN-Net: A Geometric and Neighborhood-Aware Network for Predicting Intracranial Aneurysm Rupture Risk and Assisting Clinical Decision-Making

  • 期刊论文

  • 发表日期：

    2025年12月2日

查看全文
  • Jiafeng Zhou,
  • Peiying Li,
  • Yongchang Liu,
  • Shikui Tu,
  • Bing Zhao,
  • Jieqing Wan,
  • Yongchun Chen,
  • Yunjun Yang,
  • Lei Xu

摘要

Abstract

Intracranial aneurysms bring substantial health risks due to their potential to rupture, which may lead to severe morbidity and mortality. Accurate prediction of rupture risk is essential for guiding clinical decision-making and formulating effective treatment strategies. In this study, we introduce GN-Net, a novel deep learning model designed to predict the rupture risk of intracranial aneurysms by integrating geometric and neighborhood features extracted from 3D Computed Tomography Angiography (CTA) images. GN-Net comprises two distinct branches: a geometric branch that utilizes geometric deep learning to capture the local geometric structures of aneurysms and their parent arteries, and a neighborhood-aware branch that employs 3D Convolutional Neural Networks (CNNs) and Transformer encoders to model the surrounding anatomical context. We evaluated GN-Net on an internal dataset and an external dataset. This study collected 423 valid intracranial aneurysm patients from a hospital for model training and testing. Experimental results demonstrated that our method achieved the highest prediction accuracy (91.46%), and significantly outperformed the existing state-of-the-art methods. Furthermore, the proposed model was also able to improve clinicians’ diagnostic accuracy by about 7% in assessing aneurysm rupture status. The model particularly benefited less experienced clinicians, indicating its potential to bridge expertise gaps and support decision-making processes effectively. These findings underscore GN-Net’s potential as a powerful tool for improving rupture risk prediction and supporting clinicians in making informed treatment decisions. The code for this study is available at https://github.com/YouWillLikeIt/GN-Net.

Graphical Abstract