Machine Learning-Based Prediction of Small Intracranial Aneurysm Rupture Status Using CTA-Derived Hemodynamics: A...

Area of Science:

• Neurosurgery
• Medical Imaging
• Computational Fluid Dynamics

Background:

• Small intracranial aneurysms (<5 mm) are increasingly detected.
• The rupture risk of small aneurysms is not well-understood.
• Accurate risk assessment is crucial for treatment decisions.

Purpose of the Study:

• To develop and validate machine learning models for predicting small aneurysm rupture risk.
• To integrate clinical, morphological, and hemodynamic data.
• To test model performance in external datasets.

Main Methods:

• Retrospective enrollment of 504 patients with small aneurysms.
• Development of machine learning models (SVM, random forest, logistic regression, MLP).
• Computational fluid dynamics (CFD) for hemodynamic parameter extraction.
• Internal and external validation datasets used.

Main Results:

• Support vector machine (SVM) demonstrated the best performance (AUC 0.88 training, 0.91 internal validation).
• Hemodynamic factors (stable flow, inflow patterns, flow impingement zone, OSI-CV) were key predictors.
• SVM achieved an AUC of 0.82 in external validation, comparable to internal validation (P=0.21).

Conclusions:

• Machine learning models show good performance in predicting small aneurysm rupture.
• Hemodynamic parameters are critical predictors of aneurysm rupture.
• Validated models can aid in clinical decision-making for small aneurysms.