Augmenting Prediction of Intracranial Aneurysms' Risk Status Using Velocity-Informatics: Initial Experience.

Area of Science:

• Biomedical Engineering
• Medical Imaging Analysis
• Computational Fluid Dynamics

Background:

• Intracranial aneurysms (IAs) pose a significant rupture risk.
• Accurate prediction of IA rupture status is crucial for patient management.
• Current methods may not fully capture the complex hemodynamics within IAs.

Purpose of the Study:

• To introduce and evaluate velocity-informatics for enhanced prediction of intracranial aneurysm rupture.
• To assess the efficacy of machine learning models using velocity-informatics metrics.
• To improve the accuracy of differentiating ruptured from unruptured IAs.

Main Methods:

• Generated 3D computer models from imaging data of 112 patients with anterior IAs.
• Performed computational fluid dynamics (CFD) simulations and geometrical analyses.
• Processed 3D velocity vector fields using velocity-informatics and applied four machine learning methods.

Main Results:

• Velocity-informatics metrics consistently improved prediction accuracy and area under the curve (AUC) by approximately 0.03 across four machine learning models.
• The support vector machine model with velocity-informatics achieved the highest performance: an AUC of 0.86 and 77% total accuracy.
• This approach correctly identified 60% of ruptured and 88% of unruptured IAs.

Conclusions:

• Velocity-informatics represents a promising advancement in predicting intracranial aneurysm rupture status.
• The integration of hemodynamic data through velocity-informatics enhances the predictive power of machine learning models.
• This technique offers a more accurate tool for clinical decision-making in managing patients with IAs.