Human neuroimaging research aims to improve our understanding and prediction of human behaviour and clinical phenotypes. These aims have been frustrated due to low effect sizes, with recent studies suggesting that reproducible results may require unpracticable sample sizes. Consequently, there is growing disenchantment with “one size fits all” approaches to imaging, and burgeoning interest in characterizing individual differences in brain function. Understanding the breadth of individual brain differences in health allows us to distinguish those changes that lead to psychopathology and which changes might be adaptive or protective. This symposium provides an overview of emerging methods for characterizing and comparing individual differences: individual-specific parcellations, functional alignment, and normative modelling. Individual-specific parcellations and functional alignment characterize how the same function can be located at different anatomical loci in different individuals. Modelling this aspect of individual variability improves the functional correspondence between individuals and could lead to personalized treatment targets. Normative modelling summarises the extent of human variability and identifies disease states as extremes in this natural variability. Recent advances in this field have boosted our understanding of heterogeneity and extreme deviations. We discuss the advantages and limitations of these approaches, as well as potential applications in clinical populations. This symposium will be of broad interest to researchers engaged in fMRI methods research, as well as to translational researchers who need to parse healthy variation from pathological changes.
Develop an understanding of the state-of-the-art of functional alignment and individualised parcellation methods, and their advantages and limitations in contrast to more established anatomical alignment.
Learn new methods to improve normative modelling in clinical datasets.
Understand how computational methods to parse individual differences can improve clinical predictions in psychiatric disorders
The target audience are researchers interested in brain imaging methods, particularly fMRI, and applications to clinical conditions or precision-imaging broadly.
Resting-state functional connectivity (RSFC) has shown great promise as a tool for characterizing the human brain. Until recently, most functional connectivity-based brain representations have relied on data averaged across many individuals. However, such population-level brain representation might obscure biologically meaningful individual-specific features. Here, we present a multi-session hierarchical Bayesian model to generate high-quality individual-specific parcellations. We further explored the behavioural relevance of our individual-specific parcellations and other individual-specific representations, including principal RSFC gradients, local gradients and independent components. We found that the principal gradient approach required at least 40 to 60 gradients to perform as well as parcellations and independent components. While most principal gradient studies utilize a single gradient, our results suggest that incorporating higher order gradients could provide significant behaviourally relevant information. If time permits, we will also discuss potential clinical applications of individual-specific parcellations.
, National University of Singapore Singapore, Singapore
Traditional neuroimaging methods assume that a particular brain function corresponds to the same anatomical locus across individuals. If the anatomical locus corresponding to a given function differs across individuals, then functional data will be misaligned, reducing power in group studies. Functional alignment accounts for individual differences by mapping from one individual’s brain vertices to another individual’s vertices on the basis of similar functional activations rather than anatomy alone. However, functional alignment can discard useful anatomical constraints. In this talk, I demonstrate that an interpolation between anatomical alignment and functional alignment outperforms either alignment method alone. The judicious combination of anatomical and functional information accounts for individual heterogeneity in functional topographies, increasing the predictive power of the resulting brain maps. We then combined this improved alignment technique with diffusion tractography to ask whether functional differences were linked to wiring differences. Within 97% of cortical parcels, rearrangements in individuals’ cortical functional topography predicted rearrangements of their incoming axonal fibres. These results suggest that the fine-scale topographical arrangement of axonal connectivity contributes to individuality in brain function.
Recent years have seen a rise in “naturalistic neuroimaging,” in which researchers scan participants viewing multimodal narrative stimuli—such as movies—to understand attention, memory, and emotion in dynamic contexts. While these studies have extended our understanding of how brain organization supports complex cognitive phenomena, they have also presented new challenges. In particular, these task paradigms make it especially difficult to disentangle individual differences in processing naturalistic stimuli from inter-individual variability in brain functional organization. In this talk, I will present new and emerging methods for comparing evoked brain activity across participants engaged in naturalistic tasks. In particular, I will focus on our work exploring methods for aligning participants in a high-dimensional functional space, highlighting how these functional alignment methods recover individual variability that is typically discarded as noise in group-level analyses. I will also introduce work extending these ideas with a state-space modeling approach, motivated by better capturing individual-specific variance in segmenting continuous audio-visual stimuli into discrete events.
Elizabeth DuPre, PhD
, Stanford University
Department of Psychology
Normative modelling is rapidly becoming an established method for understanding inter-individual differences in clinical and population-based cohorts. While this approach has been used principally for detecting individualised deviations from age-related trajectories, it is also ideally suited to understanding variation in arbitrary mappings between brain and behaviour, for example learning mappings between cognitive instruments and task-related brain activity, and for mapping the influence of rich clinical symptom profiles and brain structure. In this talk, I will survey some of this emerging literature, highlighting applications from our group that aim to learn multi-dimensional mappings between brain and behaviour at voxel level precision and understanding fine patterns of individual difference and how these relate to external variables, such as environmental adversities. I will also showcase methodological developments from our group that enable the identification of individuals with atypical profiles, for example based on extreme value statistics. I will show that normative modelling provides a comprehensive platform for understanding individual differences.
, Radboud University Nijmegen Nijmegen, Gelderland