Rupture risk prediction of cerebral aneurysms using a novel convolutional neural network-based deep learning model

Area of Science:

• Neurosurgery
• Medical Imaging
• Computational Fluid Dynamics

Background:

• Cerebral aneurysms pose a significant risk of rupture, leading to severe disability.
• Accurate prediction of rupture risk is crucial for timely intervention and improved patient outcomes.
• Hemodynamic factors are increasingly recognized as important indicators of aneurysm stability.

Purpose of the Study:

• To introduce a novel deep learning model for predicting cerebral aneurysm rupture risk.
• To leverage hemodynamic parameters within a convolutional neural network (CNN) framework.
• To enhance clinical decision-making regarding aneurysm treatment.

Main Methods:

• A retrospective analysis of 123 cerebral aneurysm cases was conducted.
• Hemodynamic parameters were calculated using computational fluid dynamics and fluid-structure interaction.
• A novel approach converted these parameters into images for CNN training, enhanced by new data augmentation techniques.

Main Results:

• The CNN model trained on combined wall shear stress (WSS) and strain images achieved an area under the receiver operating characteristic curve of 0.883.
• This combined model demonstrated high predictive accuracy with a sensitivity of 0.81 and specificity of 0.82.
• Individual CNNs trained on WSS or strain alone showed less robust predictive performance.

Conclusions:

• Deep learning algorithms, particularly CNNs, show promise in predicting cerebral aneurysm rupture risk.
• Integrating hemodynamic factors like WSS and strain into deep learning models offers a powerful tool for risk assessment.
• This approach can significantly improve the accuracy of rupture risk prediction, guiding clinical treatment strategies.