Recently, skeleton-based human action recognition has attracted a lot of research attention in the field of computer vision. Graph convolutional networks (GCNs), which model the human body skeletons as spatial-temporal graphs, have shown excellent results. However, the existing methods only focus on the local physical connection between the joints, and ignore the non-physical dependencies among joints. To address this issue, we propose a hypergraph neural network (Hyper-GNN) to capture both spatial-temporal information and high-order dependencies for skeleton-based action recognition. In particular, to overcome the influence of noise caused by unrelated joints, we design the Hyper-GNN to extract the local and global structure information via the hyperedge (i.e., non-physical connection) constructions. In addition, the hypergraph attention mechanism and improved residual module are induced to further obtain the discriminative feature representations. Finally, a three-stream Hyper-GNN fusion architecture is adopted in the whole framework for action recognition. The experimental results performed on two benchmark datasets demonstrate that our proposed method can achieve the best performance when compared with the state-of-the-art skeleton-based methods.

Background: Depression has been linked to disruptions in resting-state networks (RSNs). However, inconsistent findings on RSN disruptions, with variations in reported connectivity within and between RSNs, complicate the understanding of the neurobiological mechanisms underlying depression. Methods: A systematic literature search of PubMed and Web of Science identified studies that employed resting-state functional magnetic resonance imaging (fMRI) to explore RSN changes in depression. Studies using seed-based functional connectivity analysis or independent component analysis were included, and coordinate-based meta-analyses were performed to evaluate alterations in RSN connectivity both within and between networks. Results: A total of 58 studies were included, comprising 2321 patients with depression and 2197 healthy controls. The meta-analysis revealed significant alterations in RSN connectivity, both within and between networks, in patients with depression compared with healthy controls. Specifically, within-network changes included both increased and decreased connectivity in the default mode network (DMN) and increased connectivity in the frontoparietal network (FPN). Between-network findings showed increased DMN–FPN and limbic network (LN)–DMN connectivity, decreased DMN–somatomotor network and LN–FPN connectivity, and varied ventral attention network (VAN)–dorsal attentional network (DAN) connectivity. Additionally, a positive correlation was found between illness duration and increased connectivity between the VAN and DAN. Conclusions: These findings not only provide a comprehensive characterization of RSN disruptions in depression but also enhance our understanding of the neurobiological mechanisms underlying depression.

The fractional amplitude of low-frequency fluctuations (fALFF) has been widely used as potential clinical biomarkers for resting-state functional magnetic resonance imaging based schizophrenia diagnosis. However, previous studies usually measure the fALFF with specific bands from 0.01-0.08 Hz, which cannot fully delineate the complex variations of spontaneous fluctuations in the resting-state brain. As we konow, fALFF data is intrinsically constrained by the brain structure, but most of the traditional methods have not consider it in feature selection. For addressing these problems, we propose a model to classify schizophrenia in multi-frequency bands with tree-guided group sparse learning. In detail, we first acquire the fALFF data in multi-frequency bands (i.e., slow-5:0.01-0.027 Hz, slow-4:0.027-0.073 Hz, slow-3:0.073-0.198 Hz and slow-2:0.198-0.25 Hz). Then, we divide the whole brain into different candidate patches, and select those significant patches related to schizophrenia using random forest-based importancescore. Moreover, we use tree-structured sparse learning method for feature selection with above patches spatial constraint. Finally, considering biomarkers from multi-frequency bands can reflect complementary information among multiple frequency bands, we adopt the multi-kernel learning (MKL) method to combine features of multi-frequency bands for classification. Our experimental results show that these biomarkers from multi-frequency bands can achieve a classification accuracy of 91.1% on 17 schizophrenia patients and 17 healthy controls, further demonstrating the multi-frequency bands analysis can better account for classification of schizophrenia.