Comprehensive study of cortical thickness and thinning in the lifespan combining MRI, laminar architecture, and machine learning

• Ano de defesa: 2023
• Autor(a) principal: Marçal, Tamires Corrêa
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Aprendizado de máquina; Citoarquitetura; Cortical thinning; Cytoarchitecture; Envelhecimento saudável; Espessamento cortical; Healthy aging; Machine learning; Magnetic resonance imaging (MRI); Neuroplasticidade; Neuroplasticity; Poda neural; Pruning; Ressonância magnética
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/17/17163/tde-11042024-091750/

Cortical thinning is associated with pruning, neuroplasticity, and cognitive decline throughout the different phases of the lifespan. While age is a crucial factor in predicting thinning, it does not account for all its variability. To advance our comprehension of this process, we utilize Magnetic Resonance Imaging data, a Multivariate Dataset, and Machine Learning techniques. Our objective is to predict cortical thickness and thinning by analyzing a diverse set of temporal and spatial variables, including age, cortical type, lobes, brain structures, curvature, and cytoarchitectonic information. To achieve that we utilized anatomical MRI of 871 participants without a history of neurological diseases to estimate cortical thinning trajectories throughout the lifespan. We also used cytoarchitecture profiles that were estimated based on the BigBrain database. To assess the optimal method for modeling cortical thickness, we developed models based on both vertex-level and brain-structure-level. We found that the brain-structures model outperformed the vertex-level approach in predicting thickness, being able to explain 87% of its variability. To predict thinning, we began by calculating human annual cortical thinning, following which we utilized a boosting algorithm to predict thinning using three different models. A temporal model (age as only variable) achieved an r-squared of 0.79, a spatial model (all variables except age) had a score of 0.58, and temporal-spatial reached 0.84. Through the use of Shapley additive explanations in the temporal-spatial model, we see the contribution and interactions of each variable to cortical thinning. Age was the feature that most contributed to the cortical thinning, followed by layer I thickness, cortical thickness at 10y.o. and layer IV thickness. Our examination suggests that regions that experience more thinning during development tend to undergo less thinning during aging, and this correlation is linked to Layer I thickness.