Scholarlink

<p>The estimation of rupture risk in Unruptured Intracranial Aneurysm (UIA) constitutes a major area of clinical interest due to the significant morbidity and mortality rates associated with aneurysm rupture. Classic clinical models based on factors such as size and location have demonstrated limited predictive accuracy, with small aneurysms being capable of rupture and larger ones remaining stable. Recent advances in Artificial Intelligence (AI) now allow the development of more sophisticated models that integrate both geometric and hemodynamic variables, including wall shear stress (WSS) and blood flow dynamics. While previous studies have examined these factors separately, our review specifically focuses on how they are combined within AI-based predictive models for unruptured intracranial aneurysms (UIAs). This integrated approach offers a more comprehensive and patient-specific risk assessment, going beyond traditional size-based methods. A wide array of machine learning (ML) and deep learning (DL) using SVMs (Support Vector Machine) and CNNs (Convolutional Neural Network) has demonstrated much better predictive accuracy than those attained by classical methods. Minimum necessary hemodynamic parameters including WSS and oscillatory shear index (OSI) were identified as critical indicators of rupture. Moreover, the review emphasized how CFD (Computational Fluid Dynamics) merged with AI in simulating patient-specific hemodynamics, outstanding progress having been achieved in the realm of risk assessment. Currently, there are promising developments in AI models for clinical practice, but large and good-quality datasets, along with interpretation of model predictions, remain challenges. More research would further refine these models toward improvement, with increased utility in a clinical setup to better aim at patient-specific risk assessment and optimization of treatment strategies for UIAs.</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

Evaluating artificial intelligence models for rupture risk prediction in unruptured intracranial aneurysms: a focus on vessel geometry and hemodynamic insights

  • 期刊论文

  • 发表日期：

    2025年7月2日

查看全文
  • Muhammad Mohsin Khan,
  • Noman Shah,
  • Javed Iqbal,
  • Nasser M.F. El-Ghandour,
  • Miroslav Vukic,
  • Michael Lawton⁶,
  • Jacques J. Morcos,
  • Bostjan Matos,
  • Najia El-Abbadi,
  • Amir Samii,
  • Eberval Gadelha Figueiredo,
  • Franco Servadei,
  • Ahmed AlAzri,
  • Kodeeswaran M,
  • Aruni Velalakan,
  • Bipin Chaurasia⁶

摘要

The estimation of rupture risk in Unruptured Intracranial Aneurysm (UIA) constitutes a major area of clinical interest due to the significant morbidity and mortality rates associated with aneurysm rupture. Classic clinical models based on factors such as size and location have demonstrated limited predictive accuracy, with small aneurysms being capable of rupture and larger ones remaining stable. Recent advances in Artificial Intelligence (AI) now allow the development of more sophisticated models that integrate both geometric and hemodynamic variables, including wall shear stress (WSS) and blood flow dynamics. While previous studies have examined these factors separately, our review specifically focuses on how they are combined within AI-based predictive models for unruptured intracranial aneurysms (UIAs). This integrated approach offers a more comprehensive and patient-specific risk assessment, going beyond traditional size-based methods. A wide array of machine learning (ML) and deep learning (DL) using SVMs (Support Vector Machine) and CNNs (Convolutional Neural Network) has demonstrated much better predictive accuracy than those attained by classical methods. Minimum necessary hemodynamic parameters including WSS and oscillatory shear index (OSI) were identified as critical indicators of rupture. Moreover, the review emphasized how CFD (Computational Fluid Dynamics) merged with AI in simulating patient-specific hemodynamics, outstanding progress having been achieved in the realm of risk assessment. Currently, there are promising developments in AI models for clinical practice, but large and good-quality datasets, along with interpretation of model predictions, remain challenges. More research would further refine these models toward improvement, with increased utility in a clinical setup to better aim at patient-specific risk assessment and optimization of treatment strategies for UIAs.