Building a normal brain aging model with MRI imaging data
Modelling the conditional quantile of the aging process
My PhD thesis focuses on building a normal brain aging model with MRI imaging data that models the conditional quantile of the aging process. To achieve this task, statistical tools are in need to (1) understand the proportion of variance in aging that can be explained by brain morphology and to (2) validate and select from high-dimensional non-parametric quantile models for prediction purpose. My PhD work includes (a) examining the current method in estimating morphometricity with different kernels that measure anatomical similarity; (2) proposing a kernel-based lack-of-fit test for validating non-parametric quantile models. The core statistic of the lack-of-fit test will be extended to a model selection criterion in the future. Both methods are applied on UK Biobank data, to build predictive quantile models, to eventually provide a normative range of individual brain development and make clinical assessment.
People
