New morphological parameter for intracranial aneurysms and rupture risk prediction based on artificial neural networks.

Background: Numerous studies have evaluated the rupture risk of intracranial aneurysms using morphological parameters because of their good predictive capacity. However, the limitation of current morphological parameters is that they do not always allow evaluation of irregularities of intracranial aneurysms. The purpose of this study is to propose a new morphological parameter that can quantitatively describe irregularities of intracranial aneurysms and to evaluate its performance regarding rupture risk prediction.

Methods: In a retrospective study, conventional morphological parameters (aspect ratio, bottleneck ratio, height-to-width ratio, volume to ostium ratio, and size ratio) and a newly proposed morphological parameter (mass moment of inertia) were calculated for 125 intracranial aneurysms (80 unruptured and 45 ruptured aneurysms). Additionally, hemodynamic parameters (wall shear stress and strain) were calculated using computational fluid dynamics and fluid-structure interaction. Artificial neural networks trained with each parameter were used for rupture risk prediction.

Results: All components of the mass moment of inertia (I, I, and I) were significantly higher in ruptured cases than in unruptured cases (p values for I, I, and I were 0.032, 0.047, and 0.039, respectively). When the conventional morphological and hemodynamic parameters as well as the mass moment of inertia were considered together, the highest performance for rupture risk prediction was obtained (sensitivity 96.3%; specificity 85.7%; area under the receiver operating characteristic curve 0.921).

Conclusions: The mass moment of inertia would be a useful parameter for evaluating aneurysm irregularity and hence its risk of rupture. The new approach described here may help clinicians to predict the risk of aneurysm rupture more effectively.