Enhancing Power in Neuroscience: the Role of Statistical Innovation and Machine Learning

Abstract: During the last decade multisite imaging studies have become very popular due to increased statistical power and enabling the generalization of research outcomes; however, significantly different tau PET tracer properties and inter-scanner variability hinders the direct comparability of multi-scanner PET data. The tau PET imaging field is lagging in terms of statistical harmonization methods due to the complexity associated with combination of different tracers and different scanners and the lack of a gold standard. In this study we present a standardization method across PET scanners and sites in neuroimaging studies of Alzheimer's Disease using a gold standard anatomically accurate multi-chamber PET phantom. We show variability of effect sizes across different simulated groups using existing harmonization techniques. 

Integrating datasets from multiple sites has become increasingly common in various scientific fields, such as neuroimaging studies. Factor analysis is a widely used statistical method for elucidating relationships between multivariate observations and identifying the underlying factors that explain their interdependencies. However, there has been limited exploration on multi-site factor analysis that accounts for the variability and heterogeneity inherent in data collected from multiple sites or institutions. In this talk, we propose an integrative semi-confirmatory factor analysis (i-SCFA) model, that identifies shared latent factors across sites while accommodating site-specific heterogeneity in factor scores. The i-SCFA model relaxes the requirement for the prior knowledge of "non-zero loadings" in confirmatory factor analysis (CFA) by collaboratively learning the latent covariance structure across multiple sites. With its computational efficiency in identifying latent structures and providing closed-form solutions for CFA parameters, i-SCFA is particularly well-suited for high-throughput datasets. We demonstrate the empirical performance of i-SCFA through extensive simulations and apply it to a multi-site neuroimaging study. The empirical performance of i-SCFA is assessed through extensive simulations and demonstrated with the multi-site neuroimaging analysis of the Adolescent Brain Cognitive Development Study. 

Motivated by a study on adolescent mental health, a dynamic connectivity analysis is conducted using resting-state functional magnetic resonance imaging (fMRI) data. A dynamic connectivity analysis investigates how the interactions between different regions of the brain, represented by the different dimensions of a multivariate time series, change over time. Changes in the distributional properties of the data can be captured by identifying the changepoints in the time series data. The presence of changepoints in the fMRI data suggests that the connectivity between different regions of the brain changes over time. An overview of changepoint analysis and the utility of dynamic connectivity analysis is given. The novel approach for changepoint analysis that uses a mixed model framework is then described, thereby leveraging the spatial structure of the brain. The mixed model is embedded in a dynamic programming algorithm for detecting multiple changepoints in the fMRI data. The results of the proposed changepoint model in a dynamic connectivity analysis on fMRI are shown on data obtained from female adolescents, and also on data from the Adolescent Brain Cognitive Development (ABCD) Study. 

Several recent studies have raised concerns about the replicability of brain-wide association studies (BWAS). Here, we perform analyses and meta-analyses of a robust effect size index using 63 longitudinal and cross-sectional MRI studies (77,695 total scans) to demonstrate that optimizing study design is an important way to improve standardized effect sizes and replicability in BWAS. A meta-analysis of brain volume associations with age indicates that BWAS with larger covariate variance have larger effect size estimates and that the longitudinal studies we examined have systematically larger standardized effect sizes than cross-sectional studies. Analyzing age effects on global and regional brain measures in the Lifespan Brain Chart Consortium, we show that modifying longitudinal study design to increase between-subject variability and adding a single additional longitudinal measurement per subject improves effect sizes. However, evaluating these longitudinal sampling schemes on cognitive, psychopathology, and demographic associations with structural and functional brain outcome measures in the Adolescent Brain and Cognitive Development dataset shows that longitudinal studies can, counterintuitively, be detrimental to replicability. We demonstrate that the benefit of conducting longitudinal studies depends on the strengths of the between- and within-subject associations of the brain and non-brain measures. Explicitly modeling between- and within-subject effects avoids conflating the effects and allows optimizing effect sizes for them separately. These findings underscore the importance of considering design features in BWAS and emphasize that increasing sample size is not the only approach to improve the replicability of BWAS.  

The study of functional brain networks has grown tremendously over the past decade. Most functional connectivity (FC) analyses assume that FC networks are stationary across time. However, there is interest in studying changes in FC over time. Hidden Markov models (HMMs) are a useful modeling approach for FC. However, a severe limitation is that HMMs assume the sojourn time (number of consecutive time points in a state) is geometrically distributed. This encourages state switches too often. I propose a hidden semi-Markov model (HSMM) approach for inferring functional brain networks from functional magnetic resonance imaging (fMRI) data, which explicitly models the sojourn distribution. Specifically, I propose using HSMMs to find each subject's most probable series of network states, the cumulative time in each state, and the networks associated with each state. This approach is demonstrated on fMRI data from a study on older adults with obesity. Lastly, I will discuss limitations and future directions for HSMMs within state-based dynamic connectivity analysis as a whole.