The field of neuroimaging has truly become data rich, and novel analytical methods capable of gleaning meaning- ful information from large stores of imaging data are in high demand. Those methods that might also be applicable on the level of individual subjects, and thus potentially useful clinically, are of special interest. In the present study, we introduce just such a method, called nonlinear functional mapping (NFM), and demonstrate its application in the analysis of resting state fMRI (functional Magnetic Resonance Imaging) from a 242-subject subset of the IMAGEN project, a European study of adolescents that includes longitudinal phenotypic, behavioral, genetic, and neuroimaging data. NFM employs a computational technique inspired by biological evolution to discover and mathematically char- acterize interactions among ROI (regions of interest), without making linear or univariate assumptions. We show that statistics of the resulting interaction relationships comport with recent independent work, constituting a preliminary cross-validation. Furthermore, nonlinear terms are ubiquitous in the models generated by NFM, suggesting that some of the interactions characterized here are not discoverable by standard linear methods of analysis. We discuss one such nonlinear interaction in the context of a direct comparison with a procedure involving pairwise correlation, designed to be an analogous linear version of functional mapping. We find another such interaction that suggests a novel distinction in brain function between drinking and non-drinking adolescents: a tighter coupling of ROI associated with emotion, re- ward, and interoceptive processes such as thirst, among drinkers. Finally, we outline many improvements and extensions of the methodology to reduce computational expense, complement other analytical tools like graph-theoretic analysis, and allow for voxel level NFM to eliminate the necessity of ROI selection.
Abstract: Improving the accuracy of forecast models for physical systems such as the atmosphere is a crucial ongoing effort. The primary focus of recent research on these highly nonlinear systems has been errors in state estimation, but as that error has been successfully diminished, the role of model error in forecast uncertainty has duly increased. The present study is an investigation of an empirical model correction procedure involving the comparison of short forecasts with a reference 'truth' system during a training period, in order to calculate systematic (1) state-independent model bias and (2) state-dependent error patterns. An estimate of the likelihood of the latter error component is computed from the current state at every time step of model integration. The effectiveness of this technique is explored in a realistic scenario, in which the model is structurally different (in dynamics, dimension, and parametrization) from the target system. Results suggest that the correction procedure is more effective for reducing error and prolonging forecast usefulness than parameter tuning. However, the cost of this increase in average forecast accuracy is the creation of substantial qualitative differences between the dynamics of the corrected model and the true system. A method to mitigate dynamical ramifications and further increase forecast accuracy is presented.