A stochastic world model on gravity for stability inference.

The fact that objects without proper support will fall to the ground is not only a natural phenomenon, but also common sense in mind. Previous studies suggest that humans may infer objects' stability through a world model that performs mental simulations with a priori knowledge of gravity acting upon the objects. Here we measured participants' sensitivity to gravity to investigate how the world model works. We found that the world model on gravity was not a faithful replica of the physical laws, but instead encoded gravity's vertical direction as a Gaussian distribution. The world model with this stochastic feature fit nicely with participants' subjective sense of objects' stability and explained the illusion that taller objects are perceived as more likely to fall. Furthermore, a computational model with reinforcement learning revealed that the stochastic characteristic likely originated from experience-dependent comparisons between predictions formed by internal simulations and the realities observed in the external world, which illustrated the ecological advantage of stochastic representation in balancing accuracy and speed for efficient stability inference. The stochastic world model on gravity provides an example of how a priori knowledge of the physical world is implemented in mind that helps humans operate flexibly in open-ended environments.

Self-esteem mediates the relationship between the parahippocampal gyrus and decisional procrastination at resting state.

When faced with a conflict or dilemma, we tend to postpone or even avoid making a decision. This phenomenon is known as decisional procrastination. Here, we investigated the neural correlates of this phenomenon, in particular the parahippocampal gyrus (PHG) that has previously been identified in procrastination studies. In this study, we applied an individual difference approach to evaluate participants' spontaneous neural activity in the PHG and their decisional procrastination levels, assessed outside the fMRI scanner. We discovered that the fractional amplitude of low-frequency fluctuations (fALFF) in the caudal PHG (cPHG) could predict participants' level of decisional procrastination, as measured by the avoidant decision-making style. Importantly, participants' self-esteem mediated the relationship between the cPHG and decisional procrastination, suggesting that individuals with higher levels of spontaneous activity in the cPHG are likely to have higher levels of self-esteem and thus be more likely to make decisions on time. In short, our study broadens the PHG's known role in procrastination by demonstrating its link with decisional procrastination and the mediating influence of self-esteem, underscoring the need for further exploration of this mediation mechanism.

From sMRI to task-fMRI: A unified geometric deep learning framework for cross-modal brain anatomo-functional mapping.

Achieving predictions of brain functional activation patterns/task-fMRI maps from its underlying anatomy is an important yet challenging problem. Once successful, it will not only open up new ways to understand how brain anatomy influences functional organization of the brain, but also provide new technical support for the clinical use of anatomical information to guide the localization of cortical functional areas. However, due to the non-Euclidean complex architecture of brain anatomy and the inherent low signal-to-noise ratio (SNR) properties of fMRI signals, the key challenge in building such a cross-modal brain anatomo-functional mapping is how to effectively learn the context-aware information of brain anatomy and overcome the interference of noise-containing task-fMRI labels on the learning process. In this work, we propose a Unified Geometric Deep Learning framework (BrainUGDL) to perform the cross-modal brain anatomo-functional mapping task. Considering that both global and local structures of brain anatomy have an impact on brain functions from their respective perspectives, we innovatively propose the novel Global Graph Encoding (GGE) unit and Local Graph Attention (LGA) unit embedded into two parallel branches, focusing on learning the high-level global and local context information, respectively. Specifically, GGE learns the global context information of each mesh vertex by building and encoding global interactions, and LGA learns the local context information of each mesh vertex by selectively aggregating patch structure enhanced features from its spatial neighbors. The information learnt from the two branches is then fused to form a comprehensive representation of brain anatomical features for final brain function predictions. To address the inevitable measurement noise in task-fMRI labels, we further elaborate a novel uncertainty-filtered learning mechanism, which enables BrainUGDL to realize revised learning from the noise-containing labels through the estimated uncertainty. Experiments across seven open task-fMRI datasets from human connectome project (HCP) demonstrate the superiority of BrainUGDL. To our best knowledge, our proposed BrainUGDL is the first to achieve the prediction of individual task-fMRI maps solely based on brain sMRI data.

Sex dimorphic cortical brain volumes associated with antisocial behavior in young adults.

Background: Although sex differences in antisocial behavior are well-documented, the extent to which neuroanatomical differences are related to sex differences in antisocial behavior is unclear. The inconsistent results from different clinical populations exhibiting antisocial behaviors are mainly due to the heterogeneity in etiologies, comorbidity inequality, and small sample size, especially in females. Objective: The study aimed to find sexual dimorphic brain regions associated with individual differences in antisocial behavior while avoiding the issues of heterogeneity and sample size. Methods: We collected structural neuroimaging data from 281 college students (131 males, 150 females) and analyzed the data using voxel-based morphometry. Results: The gray matter volume in three brain regions correlates with self-reported antisocial behavior in males and females differently: the posterior superior temporal sulcus, middle temporal gyrus, and precuneus. The findings have controlled for the total cortical gray matter volume, age, IQ, and socioeconomic status. Additionally, we found a common neural substrate of antisocial behavior in both males and females, extending from the anterior temporal lobe to the insula. Conclusion: This is the first neuroanatomical evidence from a large non-clinical sample of young adults. The study suggests that differences in males and females in reading social cues, understanding intentions and emotions, and responding to conflicts may contribute to the modulation of brain morphometry concerning antisocial behavior.

Dissociation of categorical and coordinate spatial relations on dynamic network organization states.

Humans can flexibly represent both categorical and coordinate spatial relations. Previous research has mainly focused on hemisphere lateralization in representing these two types of spatial relations, but little is known about how distinct network organization states support representations of the two. Here we used dynamic resting-state functional connectivity (FC) to explore this question. To do this, we separated a meta-identified navigation network into a ventral and two other subnetworks. We revealed a Weak State and a Strong State within the ventral subnetwork and a Negative State and a Positive State between the ventral and other subnetworks. Further, we found the Weak State (i.e., weak but positive FC) within the ventral subnetwork was related to the ability of categorical relation recognition, suggesting that the representation of categorical spatial relations was related to weak integration among focal regions in the navigation network. In contrast, the Negative State (i.e., negative FC) between the ventral and other subnetworks was associated with the ability of coordinate relation processing, suggesting that the representation of coordinate spatial relations may require competitive interactions among widely distributed regions. In sum, our study provides the first empirical evidence revealing different focal and distributed organizations of the navigation network in representing different types of spatial information.

An fMRI dataset for whole-body somatotopic mapping in humans.

The somatotopic representation of the body is a well-established organizational principle in the human brain. Classic invasive direct electrical stimulation for somatotopic mapping cannot be used to map the whole-body topographical representation of healthy individuals. Functional magnetic resonance imaging (fMRI) has become an indispensable tool for the noninvasive investigation of somatotopic organization of the human brain using voluntary movement tasks. Unfortunately, body movements during fMRI scanning often cause large head motion artifacts. Consequently, there remains a lack of publicly accessible fMRI datasets for whole-body somatotopic mapping. Here, we present public high-resolution fMRI data to map the somatotopic organization based on motor movements in a large cohort of healthy adults (N = 62). In contrast to previous studies that were mostly designed to distinguish few body representations, most body parts are considered, including toe, ankle, leg, finger, wrist, forearm, upper arm, jaw, lip, tongue, and eyes. Moreover, the fMRI data are denoised by combining spatial independent component analysis with manual identification to clean artifacts from head motion associated with body movements.

Real-world size of objects serves as an axis of object space.

Our mind can represent various objects from physical world in an abstract and complex high-dimensional object space, with axes encoding critical features to quickly and accurately recognize objects. Among object features identified in previous neurophysiological and fMRI studies that may serve as the axes, objects' real-world size is of particular interest because it provides not only visual information for broad conceptual distinctions between objects but also ecological information for objects' affordance. Here we use deep convolutional neural networks (DCNNs), which enable direct manipulation of visual experience and units' activation, to explore how objects' real-world size is extracted to construct the axis of object space. Like the human brain, the DCNNs pre-trained for object recognition also encode objects' size as an independent axis of the object space. Further, we find that the shape of objects, rather than retinal size, context, task demands or texture features, is critical to inferring objects' size for both DCNNs and humans. In short, with DCNNs as a brain-like model, our study devises a paradigm supplemental to conventional approaches to explore the structure of object space, which provides computational support for empirical observations on human perceptual and neural representations of objects.

Development of navigation network revealed by resting-state and task-state functional connectivity.

Humans possess the essential capacity to navigate in environment, supported by multiple brain regions constituting the navigation network. Recent studies on development of the navigation network mainly examined activation changes in the medial temporal regions. It is unclear how the large-scale organization of the whole navigation network develops and whether the network organizations under resting-state and task-state develop differently. We addressed these questions by examining functional connectivity (FC) of the navigation network in 122 children (10-13 years) and 260 adults. First, we identified a modular structure in the navigation network during resting-state that included a ventral and a dorsal module. Then, we found that the intrinsic modular structure was strengthened from children to adults, that is, adults showed stronger FC within the ventral module and weaker FC between ventral and dorsal modules than children. Further, the intrinsic modular structure was loosened when performing scene-viewing task, that is, both adults and children showed decreased within-ventral FC and increased between-module FC during task- than resting-state. Finally, the task-modulated FC changes were greater in adults than in children. In sum, our study reveals age-related changes in the navigation network organization as increasing modularity under resting-state and increasing flexibility under task-state.

Semantic Relatedness Emerges in Deep Convolutional Neural Networks Designed for Object Recognition.

Human not only can effortlessly recognize objects, but also characterize object categories into semantic concepts with a nested hierarchical structure. One dominant view is that top-down conceptual guidance is necessary to form such hierarchy. Here we challenged this idea by examining whether deep convolutional neural networks (DCNNs) could learn relations among objects purely based on bottom-up perceptual experience of objects through training for object categorization. Specifically, we explored representational similarity among objects in a typical DCNN (e.g., AlexNet), and found that representations of object categories were organized in a hierarchical fashion, suggesting that the relatedness among objects emerged automatically when learning to recognize them. Critically, the emerged relatedness of objects in the DCNN was highly similar to the WordNet in human, implying that top-down conceptual guidance may not be a prerequisite for human learning the relatedness among objects. In addition, the developmental trajectory of the relatedness among objects during training revealed that the hierarchical structure was constructed in a coarse-to-fine fashion, and evolved into maturity before the establishment of object recognition ability. Finally, the fineness of the relatedness was greatly shaped by the demand of tasks that the DCNN performed, as the higher superordinate level of object classification was, the coarser the hierarchical structure of the relatedness emerged. Taken together, our study provides the first empirical evidence that semantic relatedness of objects emerged as a by-product of object recognition in DCNNs, implying that human may acquire semantic knowledge on objects without explicit top-down conceptual guidance.

DNNBrain: A Unifying Toolbox for Mapping Deep Neural Networks and Brains.

Deep neural networks (DNNs) have attained human-level performance on dozens of challenging tasks via an end-to-end deep learning strategy. Deep learning allows data representations that have multiple levels of abstraction; however, it does not explicitly provide any insights into the internal operations of DNNs. Deep learning's success is appealing to neuroscientists not only as a method for applying DNNs to model biological neural systems but also as a means of adopting concepts and methods from cognitive neuroscience to understand the internal representations of DNNs. Although general deep learning frameworks, such as PyTorch and TensorFlow, could be used to allow such cross-disciplinary investigations, the use of these frameworks typically requires high-level programming expertise and comprehensive mathematical knowledge. A toolbox specifically designed as a mechanism for cognitive neuroscientists to map both DNNs and brains is urgently needed. Here, we present DNNBrain, a Python-based toolbox designed for exploring the internal representations of DNNs as well as brains. Through the integration of DNN software packages and well-established brain imaging tools, DNNBrain provides application programming and command line interfaces for a variety of research scenarios. These include extracting DNN activation, probing and visualizing DNN representations, and mapping DNN representations onto the brain. We expect that our toolbox will accelerate scientific research by both applying DNNs to model biological neural systems and utilizing paradigms of cognitive neuroscience to unveil the black box of DNNs.

COMT-Polymorphisms Modulated Functional Profile of the Fusiform Face Area Contributes to Face-Specific Recognition Ability.

Previous studies have shown that face-specific recognition ability (FRA) is heritable; however, the neural basis of this heritability is unclear. Candidate gene studies have suggested that the catechol-O-methyltransferase (COMT) rs4680 polymorphism is related to face perception. Here, using a partial least squares (PLS) method, we examined the multivariate association between 12 genotypes of 4 COMT polymorphisms (rs6269-rs4633-rs4818-rs4680) and multimodal MRI phenotypes in the human fusiform face area (FFA), which selectively responds to face stimuli, in 338 Han Chinese adults (mean age 20.45 years; 135 males). The MRI phenotypes included gray matter volume (GMV), resting-state fractional amplitude of low-frequency fluctuations (fALFF), and face-selective blood-oxygen-level-dependent (BOLD) responses (FS). We found that the first COMT-variant component (PLS1) was positively associated with the FS but negatively associated with the fALFF in the FFA. Moreover, participants with the COMT heterozygous-HEA-haplotype showed higher PLS1 FFA-MRI scores, which were positively associated with the FRA in an old/new face recognition task, than those with the COMT homozygous HEA haplotype and HEA non-carriers, suggesting that individuals with an appropriate (intermediate) level of dopamine activity in the FFA might have better FRA. In summary, our study provides empirical evidence for the genetic and neural basis for the heritability of face recognition and informs the formation of neural module functional specificity.

Electroencephalography and Functional Magnetic Resonance Imaging-Guided Simultaneous Transcranial Direct Current Stimulation and Repetitive Transcranial Magnetic Stimulation in a Patient With Minimally Conscious State.

OBJECTIVE: A minimally conscious state (MCS) is characterized by discernible behavioral evidence of consciousness that cannot be reproduced consistently. This condition is highly challenging to treat. Recent studies have demonstrated the potential therapeutic effect of non-invasive brain stimulation in patients with MCS. In one patient in an MCS, we delivered simultaneous transcranial direct current stimulation (tDCS) and repetitive transcranial magnetic stimulation (rTMS) based on an individual brain network analysis and evaluated the therapeutic effect. METHODS: The directional transfer function (DTF) was calculated based on electroencephalograph (EEG) analysis. Global brain connectivity was calculated based on functional magnetic resonance imaging (fMRI) analysis. By referring to the EEG and fMRI results, we identified inferior parietal lobes (IPLs) as targets. In the 2-week treatment period, 14 sessions were applied to the identified bilateral parietal regions. Simultaneous 1.5-mA anodal tDCS and 5-Hz rTMS were delivered for 20 min per hemisphere in each session. Clinical evaluation scores were recorded weekly throughout the treatment. A second patient given the routine treatment was evaluated as a control. RESULTS: The clinical scores of patient 1 with MCS improved after 2 weeks of stimulation treatment, and the effect lasted for up to 1 month. EEG analysis showed a significant increase (p < 0.001) in the DTF value in the gamma band in a bilateral set of posterior regions, and fMRI showed a trend toward normalized activity in the IPLs. The clinical scores of patient 2 with coma did not improve much after 2 weeks of routine treatment. The EEG analysis showed a significant increase (p = 0.021) in the DTF value in the gamma band in a bilateral set of posterior regions. CONCLUSION: The application of EEG and fMRI to characterize the functional connectivity features of the network in an MCS patient provided a reasonable and accurate stimulation target and verified the changes in functional connectivity resulting from stimulation.

A probabilistic atlas of the human motion complex built from large-scale functional localizer data.

Accurate motion perception is critical to dealing with the changing dynamics of our visual world. A cluster known as the human MT+ complex (hMT+) has been identified as a core region involved in motion perception. Several atlases defined based on cytoarchitecture, retinotopy, connectivity, and multimodal features include homologs of the hMT+. However, an hMT+ atlas defined directly based on this region's response for motion is still lacking. Here, we identified the hMT+ based on motion responses from functional magnetic resonance imaging (fMRI) localizer data in 509 participants and then built a probabilistic atlas of the hMT+. As a result, four main findings were revealed. First, the hMT+ showed large interindividual variability across participants. Second, the atlases stabilized when the number of participants used to build the atlas was more than 100. Third, the functional hMT+ showed good agreement with the hMT+ atlases built based on cytoarchitecture, retinotopy, and connectivity, suggesting a good structural-functional correspondence. Fourth, tests on multiple fMRI data sets acquired from independent participants, imaging parameters and paradigms revealed that the functional hMT+ showed higher sensitivity than all other atlases in ROI analysis except that connectivity and multimodal hMT+ atlases in the left hemisphere could infrequently attain comparable sensitivity to the functional atlas. Taken together, our findings reveal the benefit of using large-scale functional localizer data to build a reliable and representative hMT+ atlas. Our atlas is freely available for download; it can be used to localize the hMT+ in individual participants when functional localizer data are not available.

Quantifying the variability of scene-selective regions: Interindividual, interhemispheric, and sex differences.

Scene-selective regions (SSRs), including the parahippocampal place area (PPA), retrosplenial cortex (RSC), and transverse occipital sulcus (TOS), are among the most widely characterized functional regions in the human brain. However, previous studies have mostly focused on the commonality within each SSR, providing little information on different aspects of their variability. In a large group of healthy adults (N = 202), we used functional magnetic resonance imaging to investigate different aspects of topographical and functional variability within SSRs, including interindividual, interhemispheric, and sex differences. First, the PPA, RSC, and TOS were delineated manually for each individual. We then demonstrated that SSRs showed substantial interindividual variability in both spatial topography and functional selectivity. We further identified consistent interhemispheric differences in the spatial topography of all three SSRs, but distinct interhemispheric differences in scene selectivity. Moreover, we found that all three SSRs showed stronger scene selectivity in men than in women. In summary, our work thoroughly characterized the interindividual, interhemispheric, and sex variability of the SSRs and invites future work on the origin and functional significance of these variabilities. Additionally, we constructed the first probabilistic atlases for the SSRs, which provide the detailed anatomical reference for further investigations of the scene network. Hum Brain Mapp 38:2260-2275, 2017. © 2017 Wiley Periodicals, Inc.

A test-retest dataset for assessing long-term reliability of brain morphology and resting-state brain activity.

We present a test-retest dataset for evaluation of long-term reliability of measures from structural and resting-state functional magnetic resonance imaging (sMRI and rfMRI) scans. The repeated scan dataset was collected from 61 healthy adults in two sessions using highly similar imaging parameters at an interval of 103-189 days. However, as the imaging parameters were not completely identical, the reliability estimated from this dataset shall reflect the lower bounds of the true reliability of sMRI/rfMRI measures. Furthermore, in conjunction with other test-retest datasets, our dataset may help explore the impact of different imaging parameters on reliability of sMRI/rfMRI measures, which is especially critical for assessing datasets collected from multiple centers. In addition, intelligence quotient (IQ) was measured for each participant using Raven's Advanced Progressive Matrices. The data can thus be used for purposes other than assessing reliability of sMRI/rfMRI alone. For example, data from each single session could be used to associate structural and functional measures of the brain with the IQ metrics to explore brain-IQ association.