Abstract
The dominant approach to neuroimaging data analysis employs the voxel as the unit of computation. While convenient, voxels lack biological meaning and their size is arbitrarily determined by the resolution of the image. Here, we propose a multivariate spatial model in which neuroimaging data are characterised as a linearly weighted combination of multiscale basis functions which map onto underlying brain nuclei or networks or nuclei. In this model, the elementary building blocks are derived to reflect the functional anatomy of the brain during the resting state. This model is estimated using a Bayesian framework which accurately quantifies uncertainty and automatically finds the most accurate and parsimonious combination of basis functions describing the data. We demonstrate the utility of this framework by predicting quantitative SPECT images of striatal dopamine function and we compare a variety of basis sets including generic isotropic functions, anatomical representations of the striatum derived from structural MRI, and two different soft functional parcellations of the striatum derived from resting-state fMRI (rfMRI). We found that a combination of ∼50 multiscale functional basis functions accurately represented the striatal dopamine activity, and that functional basis functions derived from an advanced parcellation technique known as Instantaneous Connectivity Parcellation (ICP) provided the most parsimonious models of dopamine function. Importantly, functional basis functions derived from resting fMRI were more accurate than both structural and generic basis sets in representing dopamine function in the striatum for a fixed model order. We demonstrate the translational validity of our framework by constructing classification models for discriminating parkinsonian disorders and their subtypes. Here, we show that ICP approach is the only basis set that performs well across all comparisons and performs better overall than the classical voxel-based approach. This spatial model constitutes an elegant alternative to voxel-based approaches in neuroimaging studies; not only are their atoms biologically informed, they are also adaptive to high resolutions, represent high dimensions efficiently, and capture long-range spatial dependencies, which are important and challenging objectives for neuroimaging data.
Highlights
Neuroimaging techniques have become invaluable tools for clinical research and practice in many brain disorders thanks to their ability to noninvasively investigate brain structure and function with relatively high spatial resolution
This plot was generated by sequentially adding basis functions to the model on the basis of their automatic relevance determination’ prior (ARD) coefficient estimated from the entire dataset
As discussed in the methods section, this provides a principled measure of the utility of each feature for predicting brain activity. These models were trained using the empirical Bayes approach similar conclusions were reached using the full Bayesian approach. This shows that (i) all higher order basis sets (ICP, independent component analysis (ICA) and bisquare) predicted the DATSCAN data extremely accurately if a sufficient number of basis functions were included in the model, whereas (ii) the anatomical basis sets were substantially less accurate across most model orders for which they were applied; (iii) the data-driven basis sets (ICP and ICA) perform better than the generic basis set across most model orders, indicating that the data-driven basis sets give rise to more parsimonious models of brain function
Summary
Neuroimaging techniques have become invaluable tools for clinical research and practice in many brain disorders thanks to their ability to noninvasively investigate brain structure and function with relatively high spatial resolution. In the classical mass-univariate approach, which entails fitting independent temporal models at each sampled brain location (i.e. each voxel), spatial dependencies are effectively disregarded or dealt with suboptimally (e.g. by smoothing the data). This ignores an important source of information encoded by statistical dependencies between brain regions. For multivariate approaches as well as mass-univariate approaches it is desirable to find parsimonious representations of brain structure or function that can more faithfully represent the underlying signal Such models may predict clinically-relevant outcomes more accurately than voxel-based approaches and may be more interpretable in the sense that discriminating features may be cleanly related to underlying neuronal units of computation
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
