Real Time Hemodynamics for Intracranial Aneurysms Practical and Fast Prediction using CFD-Supervised Deep Neural Networks

November 26, 2024

Aneurysmal subarachnoid hemorrhage is one of the earliest, deadliest, and most debilitating forms of stroke, with aneurysms often remaining asymptomatic until rupture. Overall case fatality exceeds 40%, resulting in 5–11 deaths per 1,000 people, with fewer than half of survivors living independently. The often diverging risk assessments among clinicians may be unified through hemodynamic analysis and has been studied over the past two decades using Computational Fluid Dynamics (CFD). However, the lack of standardized methodologies and extended computational times limit its adoption within clinical communities. The development of efficient and user-friendly software is necessary to facilitate the integration of this tool in such environments. This work presents the first hemodynamic predictor for intracranial aneurysms, yielding promising results that suggest potential for clinical integration soon. A standardized methodology was established to conduct 101 CFD simulations and to train supervised deep neural networks (DNNs). This approach enables implementation on standard laptops that cannot run commercial CFD software, effectively breaking down longstanding accessibility barriers. The results demonstrate that hemodynamic predictions align remarkably well with CFD results, while computation time is reduced by over 420-fold (< 2 minutes). Maximum accuracy of 87.0% for vessels similar to those in the training set and maximum model overall accuracy of 78.8% using PointNet++ was achieved with a limited sample size of 52 aneurysms and trending toward improvement with additional data.

Authors in the community:

• 

  S

  Simão Vitoriano Lourenço Nieto de Castro

  ist196489

Supervisors of this institution:

• 

  V

  Vania Silverio

  ist32734

• 

  V

  Viriato Sérgio De Almeida Semião

  ist12367

mechanical-engineering - Mechanical engineering

por - Portuguese

Instituto Superior Técnico