Automatic vascularization of massive arterial networks in the human cerebral cortex: a computational modeling approach

• Ano de defesa: 2024
• Autor(a) principal: Zilves, Eduardo Guerreiro
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Laboratório Nacional de Computação Científica
• Palavras-chave em Português: Sistema cardiovascular; Vascularização automática; Circulação cerebral humana; Hemodinâmica computacional; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://tede.lncc.br/handle/tede/395

Stroke-related diseases add up to 11% of worldwide deaths, and medical imaging techniques describe many aspects of cerebral circulation. Nevertheless, the limited resolution of these techniques hinders the assessment of smaller vessels. Computational models aid in this problem, providing tools to analyze physiological scenarios otherwise inaccessible. This work proposes a methodology to build vascular networks in the human cerebral cortex across different regions of the pial surface and the gray matter. An adaptive version of the Constrained Constructive Optimization that exploits distributed computing generates blood vessels following anatomical rules. A patient-specific geometry of the brain is used as geometric substrate, and an existing model of larger arteries is the base from where the vascular network is expanded. The geometry is partitioned into three independent territories for each large cerebral artery. A thin volume of the pial surface has blood vessels generated down to the scale of arterioles of 50 micron in lumen diameter. The network penetrates the gray matter, and vascularization is constructed by appending samples of sub-trees to each terminal vessel, which are generated separately with properties associated with the gray matter. The final network for the left hemisphere reaches 234 000 vascular segments in the pial network, and 23.3 million vascular segments including the gray matter. Vessel diameters over the pial surface vary between 2100 µm and 26 µm, and terminals between 50 µm and 60 µm. The resulting network can be coupled with existing models of the cardiovascular system to simulate pulsatile blood flow and investigate the blood pressure variations along the cerebral cortex in normal and pathological conditions. This novel approach proposes a strategy for understanding patient-specific microcirculation under different conditions, allowing to evaluate the risk of stroke, among other mechanisms involved in the onset and progress of degenerative diseases. Uma geometria do cérebro específica do paciente é usada como substrato geométrico, e um modelo existente de artérias maiores é a base de onde a rede vascular é expandida. A geometria é particionada em três territórios para cada grande artéria cerebral. Um fino volume da superfície pial possui vasos sanguíneos gerado até a escala de arteríolas de 50 mícrons de diâmetro do lúmen. A rede penetra na substância cinzenta e a vascularização é construída anexando amostras de subárvores para cada navio terminal, que são geradas separadamente com propriedades associadas com a matéria cinzenta. A rede final para o hemisfério esquerdo atinge 234.000 redes vasculares segmentos na rede pial e 23,3 milhões de segmentos vasculares incluindo a substância cinzenta.