How to extract high discriminative features that help classification from complex resting-state fMRI (rs-fMRI) data is the key to improving the accuracy of brain disease recognition such as schizophrenia. In this work, we use a weighted sparse model for brain network construction, and utilize the Kendall correlation coefficient (KCC) to extract the discriminative connectivity features for schizophrenia classification, which is conducted with the linear support vector machine. Experimental results based on the rs-fMRI of 57 schizophrenia patients and 64 healthy controls show that our proposed method is more effective (i.e., achieving a significantly higher classification accuracy, 81.82%) than other competing methods. Specifically, compared with the traditional network construction methods (Pearson’s correlation and sparse representation) and the commonly used feature selection methods (two-sample t-test and Least absolute shrinkage and selection operator (Lasso)), the algorithm proposed in this paper can more effectively extract the discriminative connectivity features between the schizophrenia patients and the healthy controls, and further improve the classification accuracy. At the same time, the discriminative connectivity features extracted in the work could be used as the potential clinical biomarkers to assist the identification of schizophrenia.
Transcranial direct current stimulation (tDCS) is an emerging non-invasive brain stimulation technique. However, the rehabilitation effect of tDCS on stroke disease is unclear. In this paper, based on electroencephalogram (EEG) and complex network analysis methods, the effect of tDCS on brain function network of stroke patients during rehabilitation was investigated. The resting state EEG signals of 31 stroke rehabilitation patients were collected and divided into stimulation group (16 cases) and control group (15 cases). The Pearson correlation coefficients were calculated between the channels, brain functional network of two groups were constructed before and after stimulation, and five characteristic parameters were analyzed and compared such as node degree, clustering coefficient, characteristic path length, global efficiency, and small world attribute. The results showed that node degree, clustering coefficient, global efficiency, and small world attributes of brain functional network in the tDCS group were significantly increased, characteristic path length was significantly reduced, and the difference was statistically significant (P < 0.05). It indicates that tDCS can improve the brain function network of stroke patients in rehabilitation period, and may provide theory and experimental basis for the application of tDCS in stroke rehabilitation treatment.
Migraine is the most common primary headache clinically, with high disability rate and heavy burden. Functional MRI (fMRI) plays a significant role in the study of migraine. This article reviews the main advances of migraine without aura (MwoA) based on resting-state fMRI in recent years, including the exploration of the mechanism of fMRI in the occurrence and development of MwoA in terms of regional functional activities and functional network connections, as well as the research progress of the potential clinical application of fMRI in aiding diagnosis and assessing treatment effect for MwoA. At last, this article summarizes the current distresses and prospects of fMRI research on MwoA.
Electroconvulsive therapy (ECT) is an interventional technique capable of highly effective neuromodulation in major depressive disorder (MDD), but its antidepressant mechanism remains unclear. By recording the resting-state electroencephalogram (RS-EEG) of 19 MDD patients before and after ECT, we analyzed the modulation effect of ECT on the resting-state brain functional network of MDD patients from multiple perspectives: estimating spontaneous EEG activity power spectral density (PSD) using Welch algorithm; constructing brain functional network based on imaginary part coherence (iCoh) and calculate functional connectivity; using minimum spanning tree theory to explore the topological characteristics of brain functional network. The results show that PSD, functional connectivity, and topology in multiple frequency bands were significantly changed after ECT in MDD patients. The results of this study reveal that ECT changes the brain activity of MDD patients, which provides an important reference in the clinical treatment and mechanism analysis of MDD.
The aim of this study is to explore the effects of continuous theta-burst transcranial magnetic stimulation (cTBS) on functional brain network in emotion processing. Before and after the intervention of cTBS over left dorsolateral prefrontal cortex (DLPFC) of ten participants who were asked to perform the emotion gender recognition task, we recorded their scalp electroencephalograms (EEG). Then we used the phase synchronization of EEG to measure the connectivity between two nodes. We then calculated the network efficiency to describe the efficiency of information transmission in brain regions. Our research showed that after the intervention of cTBS and the stimulation of the emotion face picture, there was an obvious enhancement in the event-related spectral perturbation after stimuli onset in beta band in 100–300 ms. Under the stimulation of different emotion picture, the values of global phase synchronization for negative and neutral stimuli were enhanced compared to positive ones. And the increased small-worldness was found in emotional processing. In summary, based on the effect of activity change in the left DLPFC on emotion processing brain network, the emotional processing mechanism of brain networks were preliminary explored and it provided the reference for the research of emotion processing brain network in the future.
Dementia is a neurodegenerative disease closely related to brain network dysfunction. In this study, we assessed the interdependence between brain regions in patients with early-stage dementia based on phase-lock values, and constructed a functional brain network, selecting network feature parameters for metrics based on complex network analysis methods. At the same time, the entropy information characterizing the EEG signals in time domain, frequency domain and time-frequency domain, as well as the nonlinear dynamics features such as Hjorth and Hurst indexes were extracted, respectively. Based on the statistical analysis, the feature parameters with significant differences between different conditions were screened to construct feature vectors, and finally multiple machine learning algorithms were used to realize the recognition of early categories of dementia patients. The results showed that the fusion of multiple features performed well in the categorization of Alzheimer’s disease, frontotemporal lobe dementia and healthy controls, especially in the identification of Alzheimer’s disease and healthy controls, the accuracy of β-band reached 98%, which showed its effectiveness. This study provides new ideas for the early diagnosis of dementia and computer-assisted diagnostic methods.
This study is aimed to investigate objective indicators of mental fatigue evaluation to improve the accuracy of mental fatigue evaluation. Mental fatigue was induced by a sustained cognitive task. The brain functional networks in two states (normal state and mental fatigue state) were constructed based on electroencephalogram (EEG) data. This study used complex network theory to calculate and analyze nodal characteristics parameters (degree, betweenness centrality, clustering coefficient and average path length of node), and served them as the classification features of support vector machine (SVM). Parameters of the SVM model were optimized by gird search based on 6-fold cross validation. Then, the subjects were classified. The results show that characteristic parameters of node of brain function networks can be divided into normal state and mental fatigue state, which can be used in the objective evaluation of mental fatigue state.
Brain functional network changes over time along with the process of brain development, disease, and aging. However, most of the available measurements for evaluation of the difference (or similarity) between the individual brain functional networks are for charactering static networks, which do not work with the dynamic characteristics of the brain networks that typically involve a long-span and large-scale evolution over the time. The current study proposes an index for measuring the similarity of dynamic brain networks, named as dynamic network similarity (DNS). It measures the similarity by combining the “evolutional” and “structural” properties of the dynamic network. Four sets of simulated dynamic networks with different evolutional and structural properties (varying amplitude of changes, trend of changes, distribution of connectivity strength, range of connectivity strength) were generated to validate the performance of DNS. In addition, real world imaging datasets, acquired from 13 stroke patients who were treated by transcranial direct current stimulation (tDCS), were used to further validate the proposed method and compared with the traditional similarity measurements that were developed for static network similarity. The results showed that DNS was significantly correlated with the varying amplitude of changes, trend of changes, distribution of connectivity strength and range of connectivity strength of the dynamic networks. DNS was able to appropriately measure the significant similarity of the dynamics of network changes over the time for the patients before and after the tDCS treatments. However, the traditional methods failed, which showed significantly differences between the data before and after the tDCS treatments. The experiment results demonstrate that DNS may robustly measure the similarity of evolutional and structural properties of dynamic networks. The new method appears to be superior to the traditional methods in that the new one is capable of assessing the temporal similarity of dynamic functional imaging data.
The construction of brain functional network based on resting-state functional magnetic resonance imaging (fMRI) is an effective method to reveal the mechanism of human brain operation, but the common brain functional network generally contains a lot of noise, which leads to wrong analysis results. In this paper, the least absolute shrinkage and selection operator (LASSO) model in compressed sensing is used to reconstruct the brain functional network. This model uses the sparsity of L1-norm penalty term to avoid over fitting problem. Then, it is solved by the fast iterative shrinkage-thresholding algorithm (FISTA), which updates the variables through a shrinkage threshold operation in each iteration to converge to the global optimal solution. The experimental results show that compared with other methods, this method can improve the accuracy of noise reduction and reconstruction of brain functional network to more than 98%, effectively suppress the noise, and help to better explore the function of human brain in noisy environment.