Predicting individual brain functional connectivity using a Bayesian hierarchical model.

摘要: Network-oriented analysis of functional magnetic resonance imaging (fMRI), especially resting-state fMRI, has revealed important association between abnormal connectivity and brain disorders such as schizophrenia, major depression Alzheimer's disease. Imaging-based measures have become a useful tool for investigating the pathophysiology, progression treatment response psychiatric neurodegenerative diseases. Recent studies started to explore possibility using neuroimaging help predict disease guide selection individual patients. These provide impetus develop statistical methodology that would predictive information on progression-related or treatment-related changes in neural connectivity. To this end, we propose prediction method based Bayesian hierarchical model uses individual's baseline fMRI scans, coupled with relevant subject characteristics, future A key advantage proposed is it can improve accuracy individualized by combining from both group-level patterns are common subjects similar characteristics well individual-level features particular specific subject. Furthermore, our also offers inference tools intervals quantify uncertainty variability predicted outcomes. The could be approach patient's It used post-treatment after specified regimen. Another utility applied test-retest data more reliable estimator We show there exists nice connection recently developed shrinkage community. an expectation-maximization (EM) algorithm estimation model. Simulations performed evaluate methods. illustrate application methods two examples: longitudinal ADNI2 study Kirby21 study. In simulation applications, demonstrate accurate compared alternative