Background Accumulating evidence indicates regional structural changes in the white matter (WM) of brains in patients with blepharospasm (BSP); however, whether large‐scale WM structural networks undergo widespread reorganization in these patients remains unclear. Objective We investigated topology changes and global and local features of large‐scale WM structural networks in BSP patients compared with hemifacial spasm (HFS) patients or healthy controls (HCs). Methods This cross‐sectional study applied graph theoretical analysis to assess deterministic diffusion tensor tractography findings in 41 BSP patients, 41 HFS patients, and 41 HCs. WM structural connectivity in 246 cortical and subcortical regions was assessed, and topological parameters of the resulting graphs were calculated. Networks were compared among BSP, HFS, and HCs groups. Results Compared to HCs, both BSP and HFS patients showed alterations in network integration and segregation characterized by increased global efficiency and modularity and reduced shortest path length. Moreover, increased nodal efficiency in multiple cortical and subcortical regions was found in BSP and HFS patients compared with HCs. However, these differences were not found between BSP and HFS patients. Whereas all participants showed highly similar hub distribution patterns, BSP patients had additional hub regions not present in either HFS patients or HCs, which were located in the primary head and face motor cortex and basal ganglia. Conclusions Our findings suggest that the large‐scale WM structural network undergoes an extensive reorganization in BSP, probably due to both dystonia‐specific abnormalities and facial hyperkinetic movements. © 2021 International Parkinson and Movement Disorder Society

No abstract

The threat of data-poisoning backdoor attacks on learning algorithms typically comes from the labeled data. However, in deep semi-supervised learning (SSL), unknown threats mainly stem from the unlabeled data. In this paper, we propose a novel deep hidden backdoor (DeHiB) attack scheme for SSL-based systems. In contrast to the conventional attacking methods, the DeHiB can inject malicious unlabeled training data to the semi-supervised learner so as to enable the SSL model to output premeditated results. In particular, a robust adversarial perturbation generator regularized by a unified objective function is proposed to generate poisoned data. To alleviate the negative impact of the trigger patterns on model accuracy and improve the attack success rate, a novel contrastive data poisoning strategy is designed. Using the proposed data poisoning scheme, one can implant the backdoor into the SSL model using the raw data without hand-crafted labels. Extensive experiments based on CIFAR10 and CIFAR100 datasets demonstrated the effectiveness and crypticity of the proposed scheme.

No abstract

Blepharospasm is traditionally thought to be a movement disorder that results from basal ganglia dysfunction. Recently, accumulating morphometric studies have revealed structural alterations outside the basal ganglia, such as in the brainstem, cerebellum, and sensorimotor cortex, suggesting that blepharospasm may result from network disorders. However, the temporal and causal relationships between structural alterations, and whether there are disease duration-related hierarchical structural changes in these patients remain largely unknown. Structural magnetic resonance imaging was performed in 62 patients with blepharospasm, 62 patients with hemifacial spasm, and 62 healthy controls to assess the structural alterations using voxel-based morphology and structural covariance networks. The use of the causal structural covariance network, modularity analysis, and functional decoding were subsequently performed to map the causal effect of gray matter change pattern, hierarchical topography, and functional characterizations of the structural network throughout the disease duration of blepharospasm. Greater gray matter volume in the left and right supplementary motor areas was identified in patients with blepharospasm compared to that in patients with hemifacial spasm and healthy controls, whereas no significant difference was identified between patients with hemifacial spasm and healthy controls. In addition, increased gray matter volume covariance between the right supplementary motor area and right brainstem, left superior frontal gyrus, left supplementary motor area, and left paracentral gyrus was found in patients with blepharospasm compared to healthy controls. Further causal structural covariance network, modularity analysis, and functional decoding showed that the right supplementary motor area served as a driving core in patients with blepharospasm, extending greater gray matter volume to areas in the cortico-basal ganglia-brainstem motor pathway and cortical regions in the vision-motor integration pathway. Taken together, our results suggest that the right supplementary motor area is an early and important pathologically impaired region in patients with blepharospasm. With a longer duration of blepharospasm, increased gray matter volume extends from the right supplementary motor area to the cortico-basal ganglia motor and visual-motor integration pathways, showing a hierarchy of structural abnormalities in the disease progression of blepharospasm, which provides novel evidence to support the notion that blepharospasm may arise from network disorders and is associated with a wide range of gray matter abnormalities.