Abstract:Multivariate machine learning methods are increasingly used to analyze neuroimaging data, often replacing more traditional "mass univariate" techniques that fit data one voxel at a time. In the functional magnetic resonance imaging (fMRI) literature, this has led to broad application of "off-the-shelf" classification and regression methods. These generic approaches allow investigators to use ready-made algorithms to accurately decode perceptual, cognitive, or behavioral states from distributed patterns of neural activity. However, when applied to correlated whole-brain fMRI data these methods suffer from coefficient instability, are sensitive to outliers, and yield dense solutions that are hard to interpret without arbitrary thresholding. Here, we develop variants of the the Graph-constrained Elastic Net (GraphNet), ..., we (1) extend GraphNet to include robust loss functions that confer insensitivity to outliers, (2) equip them with "adaptive" penalties that asymptotically guarantee correct variable selection, and (3) develop a novel sparse structured Support Vector GraphNet classifier (SVGN). When applied to previously published data, these efficient whole-brain methods significantly improved classification accuracy over previously reported VOI-based analyses on the same data while discovering task-related regions not documented in the original VOI approach. Critically, GraphNet estimates generalize well to out-of-sample data collected more than three years later on the same task but with different subjects and stimuli. By enabling robust and efficient selection of important voxels from whole-brain data taken over multiple time points (>100,000 "features"), these methods enable data-driven selection of brain areas that accurately predict single-trial behavior within and across individuals.

Simultaneous Inference in Multiple Matrix-Variate Graphs for High-Dimensional Neural Recordings

Nonparametric Covariance Regression for Massive Neural Data on Restricted Covariates via Graph

Simultaneous inference for a high-dimensional precision matrix

High-dimensional multisubject time series transition matrix inference with application to brain connectivity analysis

A Deconvolution-Based Approach to Identifying Large-Scale Effective Connectivity.

Hypothesis Testing of Matrix Graph Model with Application to Brain Connectivity Analysis.

Whole-brain Prediction Analysis with GraphNet

A Constrained, Weighted-L1 Minimization Approach for Joint Discovery of Heterogeneous Neural Connectivity Graphs

Inference of Dynamic Graph Changes for Functional Connectome

Estimating and Inferring the Maximum Degree of Stimulus-Locked Time-Varying Brain Connectivity Networks

Joint analysis of spikes and local field potentials using copula

Large-scale Multiple Testing of Cross-covariance Functions with Applications to Functional Network Models

Joint Inference of Multiple Graphs from Matrix Polynomials

Selective Inference for Sparse Multitask Regression with Applications in Neuroimaging

Simultaneous Cluster Structure Learning and Estimation of Heterogeneous Graphs for Matrix‐variate Fmri Data

Inferring functional relations from synthetic fMRI data using large-scale Nonlinear Granger Causality (lsNGC)

Modeling Brain Connectivity with Graphical Models on Frequency Domain

Testing for Differences in Gaussian Graphical Models: Applications to Brain Connectivity

Joint learning of multiple Granger causal networks via non-convex regularizations: Inference of group-level brain connectivity

Model-Based Bayesian Inference of Neural Activity and Connectivity from All-Optical Interrogation of A Neural Circuit

Scalable Bayesian Functional Connectivity Inference for Multi-Electrode Array Recordings