Machine learning techniques have gained prominence for the analysis of resting-state functional Magnetic Resonance Imaging (rs-fMRI) data. Here, we present an overview of various unsupervised and supervised machine learning applications to rs-fMRI. We present a methodical taxonomy of machine learning methods in resting-state fMRI. We identify three major divisions of unsupervised learning methods with regard to their applications to rs-fMRI, based on whether they discover principal modes of variation across space, time or population. Next, we survey the algorithms and rs-fMRI feature representations that have driven the success of supervised subject-level predictions. The goal is to provide a high-level overview of the burgeoning field of rs-fMRI from the perspective of machine learning applications.

Conscious perception occurs within less than 1 s. To study events on this time scale we used direct electrical recordings from the human cerebral cortex during a conscious visual perception task. Faces were presented at individually titrated visual threshold for 9 subjects while measuring broadband 40-115 Hz gamma power in a total of 1621 intracranial electrodes widely distributed in both hemispheres. Surface maps and k-means clustering analysis showed initial activation of visual cortex for both perceived and non-perceived stimuli. However, only stimuli reported as perceived then elicited a forward-sweeping wave of activity throughout the cerebral cortex accompanied by large-scale network switching. Specifically, a monophasic wave of broadband gamma activation moves through bilateral association cortex at a rate of approximately 150 mm/s and eventually reenters visual cortex for perceived but not for non-perceived stimuli. Meanwhile, the default mode network and the initial visual cortex and higher association cortex networks are switched off for the duration of conscious stimulus processing. Based on these findings, we propose a new "switch-and-wave" model for the processing of consciously perceived stimuli. These findings are important for understanding normal conscious perception and may also shed light on its vulnerability to disruption by brain disorders.

The specificty and sensitivity of resting state functional MRI (rs-fMRI) measurements depend on preprocessing choices, such as the parcellation scheme used to define regions of interest (ROIs). In this study, we critically evaluate the effect of brain parcellations on machine learning models applied to rs-fMRI data. Our experiments reveal an intriguing trend: On average, models with stochastic parcellations consistently perform as well as models with widely used atlases at the same spatial scale. We thus propose an ensemble learning strategy to combine the predictions from models trained on connectivity data extracted using different (e.g., stochastic) parcellations. We further present an implementation of our ensemble learning strategy with a novel 3D Convolutional Neural Network (CNN) approach. The proposed CNN approach takes advantage of the full-resolution 3D spatial structure of rs-fMRI data and fits non-linear predictive models. Our ensemble CNN framework overcomes the limitations of traditional machine learning models for connectomes that often rely on region-based summary statistics and/or linear models. We showcase our approach on a classification (autism patients versus healthy controls) and a regression problem (prediction of subject's age), and report promising results.

No abstract