Brain spontaneous fluctuations in sensorimotor regions were directly related to eyes open and eyes closed: evidences from a machine learning approach

Abstract

Previous studies have demonstrated that the difference between resting-state brain activations depends on whether the subject was eyes open (EO) or eyes closed (EC). However, whether the spontaneous fluctuations are directly related to these two different resting states are still largely unclear. In the present study, we acquired resting-state functional magnetic resonance imaging data from 24 healthy subjects (11 males, 20.17 ± 2.74 years) under the EO and EC states. The amplitude of the spontaneous brain activity in low-frequency band was subsequently investigated by using the metric of fractional amplitude of low frequency fluctuation (fALFF) for each subject under each state. A support vector machine (SVM) analysis was then applied to evaluate whether the category of resting states could be determined from the brain spontaneous fluctuations. We demonstrated that these two resting states could be decoded from the identified pattern of brain spontaneous fluctuations, predominantly based on fALFF in the sensorimotor module. Specifically, we observed prominent relationships between increased fALFF for EC and decreased fALFF for EO in sensorimotor regions. Overall, the present results indicate that a SVM performs well in the discrimination between the brain spontaneous fluctuations of distinct resting states and provide new insight into the neural substrate of the resting states during EC and EO.

Partial covariance based functional connectivity computation using Ledoit–Wolf covariance regularization

“…Functional connectivity is also affected. Consistent with previous literature we found broad changes in full correlation functional connectivity (Figure 5) (Jao et al, 2013, Liang et al, 2014, Xu et al, 2014). In particular, we observed an increasingly strong negative correlation between visual cortices and the thalamus (Zou et al, 2009, McAvoy et al, 2012).…”

Decoding visual network‐related dynamic functional connectivity for eyes‐open and eyes‐closed using machine learning

“…3,4 Specially, the regional fALFF exhibited discriminative power in sensorimotor areas for classification of EO/EC. 15 In addition to distinctive sensorimotor-related FC, the visual and auditory network-related FC was also different between EO/EC, implying the characteristic multisensory activity might play important roles in interoceptive imagination during EC state. 1,15 In this study, the visual networks-related dynamic FC exhibited discriminative power in classification tasks.…”

“…In addition to voxel-wise measures, regional fractional ALFF (fALFF) were applied to classify EO/EC states, and achieved total accuracy of 85%. 15 The effective connectivity between ICNs could discriminate EO/EC states with total accuracy of 97%. 10 However, the results were cross-validated using a small number of samples.…”

“…Machine learning approaches have been applied to build discriminative models for EO/EC. 10,[14][15][16] The goals of discriminative models for EC/EC identification are twofold: to discover the neural mechanisms for EO/EC using data mining and to classify the states of EO/EC using machine learning. On the one hand, revealing the discriminative patterns under EO/EC using resting state fMRI could help better understanding the neural mechanisms of directional visual attention.…”

“…14,16 In regional level, the sensorimotor regions-related fractional ALFF (fALFF) could classify EO/EC states using SVM. 15 However, the regional measures represented limited information within the region of interests (ROIs). In network level, the directional conditional dependencies between intrinsic connectivity networks (ICNs) could discriminate EO/EC states through Gaussian Bayesian network learning and SVM.…”