Novel Integrated System for Segmentation and Analysis of Intracranial Arteries and Aneurysms Using Convolutional Neural Networks and Computational Fluid Dynamics

Overview

Brain Aneurysms are a significant challenge for neurosurgeons due to the often -fatal consequences of their rupture. Therefore, it is crucial to develop methods that enable doctors to detect brain aneurysms from Magnetic resonance angiography (MRA) images. Additionally, providing tools for assessing rupture risk will aid them in making informed decisions about surgical interventions. However, currently available tools do not adequately assist physicians in identifying and assessing aneurysm risks. This study proposes an AI model that segments brain MRA images to obtain both the artery and aneurysm masks. I have trained several state -of-the-art medical image segmentation models for this task. Using the Aneurysm mask, useful geometrical features can be extracted and analyzed using persistent homology to assist the neurosurgeon when performing surgery. For effective Computer Aided Engineering process, I start with converting these 3D images to Computer-Aided Design (CAD) models for blood flow pattern with final step of analysis using Computational Fluid Dynamics (CFD). I use the CFD simulations to investigate hemodynamic parameters such as blood flow and velocity. This modeling and simulation help to perform early prediction on the geometrical effects of hemodynamics. Preliminary results suggest that this novel approach can significantly enhance physicians' ability to assess aneurysm risks accurately and make informed decisions about surgical interventions. This stu dy contributes to the advancement of neurosurgical diagnostics and paves the way for more effective and personalized treatment strategies for brain aneurysms. Mississippi