Identifying Brain Regions Related to Word Prediction During Listening to Japanese Speech by Combining a LSTM Language Model and MEG

Recently, a neuroscientific approach has revealed that humans understand language while subconsciously predicting the next word from the preceding context. Most studies on human word prediction have investigated the correlations between brain activity while reading or listening to sentences on functional magnetic resonance imaging (fMRI) and the predictive difficulty of each word in a sentence calculated by the N-gram language model. However, because of its low temporal resolution, fMRI is not optimal for identifying the changes in brain activity that accompany language comprehension. In addition, the N-gram language model is a simple computational structure that does not account for the structure of the human brain. Furthermore, it is necessary for humans to retain information prior to the N-1 word in order to form a contextual understanding of a presented story. Therefore, in the present study, we measured brain activity using magnetoencephalography (MEG), which has a higher temporal resolution than fMRI, and calculated the prediction difficulty of words using a long short-term memory language model (LSTMLM), which is based on a neural network inspired by the structure of the human brain and has longer information retention than the N-gram language model. We then identified the brain regions involved in language prediction during Japanese-language speech listening using encoding and decoding analyses. In addition to surprisal-related regions revealed in previous studies, such as the superior temporal gyrus, fusiform gyrus, and temporal pole, we also found relationships between surprisal and brain activity in other regions, including the insula, superior temporal sulcus, and middle temporal gyrus, which are believed to be involved in longer-term, sentence-level cognitive processing.

Download Full-text

Distinguishing hypochondriasis and schizophrenia using regional homogeneity: a resting-state fMRI study and support vector machine analysis

Acta Neuropsychiatrica ◽

10.1017/neu.2021.9 ◽

2021 ◽

pp. 1-29

Author(s):

Kangyu Jin ◽

Zhe Shen ◽

Guoxun Feng ◽

Zhiyong Zhao ◽

Jing Lu ◽

...

Keyword(s):

Cognitive Function ◽

Resting State ◽

Parietal Lobe ◽

Brain Activity ◽

Resting State Fmri ◽

Brain Regions ◽

Regional Homogeneity ◽

Middle Temporal Gyrus ◽

Support Vector ◽

Fmri Study

Abstract Objective: A few former studies suggested there are partial overlaps in abnormal brain structure and cognitive function between Hypochondriasis (HS) and schizophrenia (SZ). But their differences in brain activity and cognitive function were unclear. Methods: 21 HS patients, 23 SZ patients, and 24 healthy controls (HC) underwent Resting-state functional magnetic resonance imaging (rs-fMRI) with the regional homogeneity analysis (ReHo), subsequently exploring the relationship between ReHo value and cognitive functions. The support vector machines (SVM) were used on effectiveness evaluation of ReHo for differentiating HS from SZ. Results: Compared with HC, HS showed significantly increased ReHo values in right middle temporal gyrus (MTG), left inferior parietal lobe (IPL) and right fusiform gyrus (FG), while SZ showed increased ReHo in left insula, decreased ReHo values in right paracentral lobule. Additionally, HS showed significantly higher ReHo values in FG, MTG and left paracentral lobule but lower in insula than SZ. The higher ReHo values in insula were associated with worse performance in MCCB in HS group. SVM analysis showed a combination of the ReHo values in insula and FG was able to satisfactorily distinguish the HS and SZ patients. Conclusion: our results suggested the altered default mode network (DMN), of which abnormal spontaneous neural activity occurs in multiple brain regions, might play a key role in the pathogenesis of HS, and the resting-state alterations of insula closely related to cognitive dysfunction in HS. Furthermore, the combination of the ReHo in FG and insula was a relatively ideal indicator to distinguish HS from SZ.

Download Full-text

Quantifying differences between passive and task-evoked intrinsic functional connectivity in a large-scale brain simulation

10.1101/250894 ◽

2018 ◽

Author(s):

Antonio Ulloa ◽

Barry Horwitz

Keyword(s):

Computational Model ◽

Resting State ◽

Short Term Memory ◽

Brain Activity ◽

Brain Regions ◽

Intrinsic Activity ◽

Task Execution ◽

Short Term ◽

Visual Short Term Memory ◽

And Task

AbstractEstablishing a connection between intrinsic and task-evoked brain activity is critical because it would provide a way to map task-related brain regions in patients unable to comply with such tasks. A crucial question within this realm is to what extent the execution of a cognitive task affects the intrinsic activity of brain regions not involved in the task. Computational models can be useful to answer this question because they allow us to distinguish task from non-task neural elements while giving us the effects of task execution on non-task regions of interest at the neuroimaging level. The quantification of those effects in a computational model would represent a step towards elucidating the intrinsic versus task-evoked connection. Here we used computational modeling and graph theoretical metrics to quantify changes in intrinsic functional brain connectivity due to task execution. We used our Large-Scale Neural Modeling framework to embed a computational model of visual short-term memory into an empirically derived connectome. We simulated a neuroimaging study consisting of ten subjects performing passive fixation (PF), passive viewing (PV) and delay match-to-sample (DMS) tasks. We used the simulated BOLD fMRI time-series to calculate functional connectivity (FC) matrices and used those matrices to compute several graph theoretical measures. After determining that the simulated graph theoretical measures were largely consistent with experiments, we were able to quantify the differences between the graph metrics of the PF condition and those of the PV and DMS conditions. Thus, we show that we can use graph theoretical methods applied to simulated brain networks to aid in the quantification of changes in intrinsic brain functional connectivity during task execution. Our results represent a step towards establishing a connection between intrinsic and task-related brain activity.Author SummaryStudies of resting-state conditions are popular in neuroimaging. Participants in resting-state studies are instructed to fixate on a neutral image or to close their eyes. This type of study has advantages over traditional task-based studies, including its ability to allow participation of those with difficulties performing tasks. Further, a resting-state neuroimaging study reveals intrinsic activity of participants’ brains. However, task-related brain activity may change this intrinsic activity, much as a stone thrown in a lake causes ripples on the water’s surface. Can we measure those activity changes? To answer that question, we merged a computational model of visual short-term memory (task regions) with an anatomical model incorporating major connections between brain regions (non-task regions). In a computational model, unlike real data, we know how different regions are connected and which regions are doing the task. First, we simulated neuronal and neuroimaging activity of both task and non-task regions during three conditions: passive fixation (baseline), passive viewing, and visual short-term memory. Then, applying graph theory to the simulated neuroimaging of non-task regions, we computed differences between the baseline and the other conditions. Our results show that we can measure changes in non-task regions due to brain activity changes in task-related regions.

Download Full-text

The neural bases of program comprehension: a coordinate-based fMRI meta-analysis

10.1101/2021.04.15.439937 ◽

2021 ◽

Author(s):

Yoshiharu Ikutani ◽

Takeshi D. Itoh ◽

Takatomi Kubo

Keyword(s):

Brain Activity ◽

Meta Analysis ◽

Inferior Frontal Gyrus ◽

Posterior Part ◽

Program Comprehension ◽

Brain Regions ◽

Middle Temporal Gyrus ◽

Activation Patterns ◽

Neural Bases ◽

Kernel Density Analysis

AbstractThe understanding of brain activity during program comprehension have advanced thanks to noninvasive neuroimaging techniques, such as functional magnetic resonance imaging (fMRI). However, individual neuroimaging studies of program comprehension often provided inconsistent results and made it difficult to identify the neural bases. To identify the essential brain regions, this study performed a small meta-analysis on recent fMRI studies of program comprehension using multilevel kernel density analysis (MKDA). Our analysis identified a set of brain regions consistently activated in various program comprehension tasks. These regions consisted of three clusters, each of which centered at the left inferior frontal gyrus pars triangularis (IFG Tri), posterior part of middle temporal gyrus (pMTG), and right middle frontal gyrus (MFG). Additionally, subsequent analyses revealed relationships among the activation patterns in the previous studies and multiple cognitive functions. These findings suggest that program comprehension mainly recycles the language-related networks and partially employs other domain-general resources in the human brain.

Download Full-text

Parental education is associated with differential engagement of neural pathways during inhibitory control

Scientific Reports ◽

10.1038/s41598-021-04152-4 ◽

2022 ◽

Vol 12 (1) ◽

Author(s):

Christopher N. Cascio ◽

Nina Lauharatanahirun ◽

Gwendolyn M. Lawson ◽

Martha J. Farah ◽

Emily B. Falk

Keyword(s):

Socioeconomic Status ◽

Response Inhibition ◽

Parental Education ◽

Brain Activity ◽

Inferior Frontal Gyrus ◽

Brain Regions ◽

Middle Temporal Gyrus ◽

Neural Pathways ◽

Performance Differences ◽

Education Performance

AbstractResponse inhibition and socioeconomic status (SES) are critical predictors of many important outcomes, including educational attainment and health. The current study extends our understanding of SES and cognition by examining brain activity associated with response inhibition, during the key developmental period of adolescence. Adolescent males (N = 81), aged 16–17, completed a response inhibition task while undergoing fMRI brain imaging and reported on their parents’ education, one component of socioeconomic status. A region of interest analysis showed that parental education was associated with brain activation differences in the classic response inhibition network (right inferior frontal gyrus + subthalamic nucleus + globus pallidus) despite the absence of consistent parental education-performance effects. Further, although activity in our main regions of interest was not associated with performance differences, several regions that were associated with better inhibitory performance (ventromedial prefrontal cortex, middle frontal gyrus, middle temporal gyrus, amygdala/hippocampus) also differed in their levels of activation according to parental education. Taken together, these results suggest that individuals from households with higher versus lower parental education engage key brain regions involved in response inhibition to differing degrees, though these differences may not translate into performance differences.

Download Full-text

Default-Mode Activity during a Passive Sensory Task: Uncoupled from Deactivation but Impacting Activation

Journal of Cognitive Neuroscience ◽

10.1162/0898929042568532 ◽

2004 ◽

Vol 16 (9) ◽

pp. 1484-1492 ◽

Cited By ~ 466

Author(s):

Michael D. Greicius ◽

Vinod Menon

Keyword(s):

Episodic Memory ◽

Cognitive Processing ◽

Neural Activity ◽

Default Mode Network ◽

Brain Activity ◽

Block Design ◽

Brain Regions ◽

Posterior Cingulate Cortex ◽

High Demand ◽

Default Mode

Deactivation refers to increased neural activity during low-demand tasks or rest compared with high-demand tasks. Several groups have reported that a particular set of brain regions, including the posterior cingulate cortex and the medial prefrontal cortex, among others, is consistently deactivated. Taken together, these typically deactivated brain regions appear to constitute a default-mode network of brain activity that predominates in the absence of a demanding external task. Examining a passive, block-design sensory task with a standard deactivation analysis (rest epochs vs. stimulus epochs), we demonstrate that the default-mode network is undetectable in one run and only partially detectable in a second run. Using independent component analysis, however, we were able to detect the full default-mode network in both runs and to demonstrate that, in the majority of subjects, it persisted across both rest and stimulus epochs, uncoupled from the task waveform, and so mostly undetectable as deactivation. We also replicate an earlier finding that the default-mode network includes the hippocampus suggesting that episodic memory is incorporated in default-mode cognitive processing. Furthermore, we show that the more a subject's default-mode activity was correlated with the rest epochs (and “deactivated” during stimulus epochs), the greater that subject's activation to the visual and auditory stimuli. We conclude that activity in the default-mode network may persist through both experimental and rest epochs if the experiment is not sufficiently challenging. Time-series analysis of default-mode activity provides a measure of the degree to which a task engages a subject and whether it is sufficient to interrupt the processes—presumably cognitive, internally generated, and involving episodic memory—mediated by the default-mode network.

Download Full-text

DS-GCNs: Connectome Classification using Dynamic Spectral Graph Convolution Networks with Assistant Task Training

Cerebral Cortex ◽

10.1093/cercor/bhaa292 ◽

2020 ◽

Author(s):

Xiaodan Xing ◽

Qingfeng Li ◽

Mengya Yuan ◽

Hao Wei ◽

Zhong Xue ◽

...

Keyword(s):

Short Term Memory ◽

Brain Activity ◽

Brain Regions ◽

Brain Network ◽

Disease Classification ◽

Sliding Windows ◽

Convolutional Networks ◽

Gender And Age ◽

The Brain ◽

Normal Controls

Abstract Functional connectivity (FC) matrices measure the regional interactions in the brain and have been widely used in neurological brain disease classification. A brain network, also named as connectome, could form a graph structure naturally, the nodes of which are brain regions and the edges are interregional connectivity. Thus, in this study, we proposed novel graph convolutional networks (GCNs) to extract efficient disease-related features from FC matrices. Considering the time-dependent nature of brain activity, we computed dynamic FC matrices with sliding windows and implemented a graph convolution–based LSTM (long short–term memory) layer to process dynamic graphs. Moreover, the demographics of patients were also used as additional outputs to guide the classification. In this paper, we proposed to utilize the demographic information as extra outputs and to share parameters among three networks predicting subject status, gender, and age, which serve as assistant tasks. We tested the performance of the proposed architecture in ADNI II dataset to classify Alzheimer’s disease patients from normal controls. The classification accuracy, sensitivity, and specificity reach 90.0%, 91.7%, and 88.6%, respectively, on ADNI II dataset.

Download Full-text

FR-Match: Robust matching of cell type clusters from single cell RNA sequencing data using the Friedman-Rafsky non-parametric test

10.1101/2020.05.01.073445 ◽

2020 ◽

Author(s):

Yun Zhang ◽

Brian D. Aevermann ◽

Trygve E. Bakken ◽

Jeremy A. Miller ◽

Rebecca D. Hodge ◽

...

Keyword(s):

Human Brain ◽

Single Cell ◽

Rna Sequencing ◽

Cell Types ◽

R Package ◽

Brain Regions ◽

Cortical Layer ◽

Middle Temporal Gyrus ◽

Cell Type ◽

Sequencing Data

AbstractSingle cell/nucleus RNA sequencing (scRNAseq) is emerging as an essential tool to unravel the phenotypic heterogeneity of cells in complex biological systems. While computational methods for scRNAseq cell type clustering have advanced, the ability to integrate datasets to identify common and novel cell types across experiments remains a challenge. Here, we introduce a cluster-to-cluster cell type matching method – FR-Match – that utilizes supervised feature selection for dimensionality reduction and incorporates shared information among cells to determine whether two cell type clusters share the same underlying multivariate gene expression distribution. FR-Match is benchmarked with existing cell-to-cell and cell-to-cluster cell type matching methods using both simulated and real scRNAseq data. FR-Match proved to be a stringent method that produced fewer erroneous matches of distinct cell subtypes and had the unique ability to identify novel cell phenotypes in new datasets. In silico validation demonstrated that the proposed workflow is the only self-contained algorithm that was robust to increasing numbers of true negatives (i.e. non-represented cell types). FR-Match was applied to two human brain scRNAseq datasets sampled from cortical layer 1 and full thickness middle temporal gyrus. When mapping cell types identified in specimens isolated from these overlapping human brain regions, FR-Match precisely recapitulated the laminar characteristics of matched cell type clusters, reflecting their distinct neuroanatomical distributions. An R package and Shiny application are provided at https://github.com/JCVenterInstitute/FRmatch for users to interactively explore and match scRNAseq cell type clusters with complementary visualization tools.

Download Full-text

Altered Spontaneous Neural Activity and Functional Connectivity in Parkinson’s Disease With Subthalamic Microlesion

Frontiers in Neuroscience ◽

10.3389/fnins.2021.699010 ◽

2021 ◽

Vol 15 ◽

Author(s):

Bei Luo ◽

Yue Lu ◽

Chang Qiu ◽

Wenwen Dong ◽

Chen Xue ◽

...

Keyword(s):

Parkinson’S Disease ◽

Parkinson's Disease ◽

Functional Connectivity ◽

Brain Activity ◽

Inferior Frontal Gyrus ◽

Low Frequency ◽

Brain Regions ◽

Angular Gyrus ◽

Middle Temporal Gyrus ◽

Precentral Gyrus

BackgroundTransient improvement in motor symptoms are immediately observed in patients with Parkinson’s disease (PD) after an electrode has been implanted into the subthalamic nucleus (STN) for deep brain stimulation (DBS). This phenomenon is known as the microlesion effect (MLE). However, the underlying mechanisms of MLE is poorly understood.PurposeWe utilized resting state functional MRI (rs-fMRI) to evaluate changes in spontaneous brain activity and networks in PD patients during the microlesion period after DBS.MethodOverall, 37 PD patients and 13 gender- and age-matched healthy controls (HCs) were recruited for this study. Rs-MRI information was collected from PD patients three days before DBS and one day after DBS, whereas the HCs group was scanned once. We utilized the amplitude of low-frequency fluctuation (ALFF) method in order to analyze differences in spontaneous whole-brain activity among all subjects. Furthermore, functional connectivity (FC) was applied to investigate connections between other brain regions and brain areas with significantly different ALFF before and after surgery in PD patients.ResultRelative to the PD-Pre-DBS group, the PD-Post-DBS group had higher ALFF in the right putamen, right inferior frontal gyrus, right precentral gyrus and lower ALFF in right angular gyrus, right precuneus, right posterior cingulate gyrus (PCC), left insula, left middle temporal gyrus (MTG), bilateral middle frontal gyrus and bilateral superior frontal gyrus (dorsolateral). Functional connectivity analysis revealed that these brain regions with significantly different ALFF scores demonstrated abnormal FC, largely in the temporal, prefrontal cortices and default mode network (DMN).ConclusionThe subthalamic microlesion caused by DBS in PD was found to not only improve the activity of the basal ganglia-thalamocortical circuit, but also reduce the activity of the DMN and executive control network (ECN) related brain regions. Results from this study provide new insights into the mechanism of MLE.

Download Full-text

Estimation and Validation of Individualized Dynamic Brain Models with Resting State fMRI

10.1101/678243 ◽

2019 ◽

Cited By ~ 1

Author(s):

Matthew F. Singh ◽

Todd S. Braver ◽

Michael W. Cole ◽

ShiNung Ching

Keyword(s):

Human Brain ◽

Resting State ◽

Nonlinear Dynamical Systems ◽

Brain Activity ◽

Network Models ◽

Resting State Fmri ◽

Brain Regions ◽

Data Driven ◽

Imaging Data ◽

Individual Level

AbstractA key challenge for neuroscience is to develop generative, causal models of the human nervous system in an individualized, data-driven manner. Previous initiatives have either constructed biologically-plausible models that are not constrained by individual-level human brain activity or used data-driven statistical characterizations of individuals that are not mechanistic. We aim to bridge this gap through the development of a new modeling approach termed Mesoscale Individualized Neurodynamic (MINDy) modeling, wherein we fit nonlinear dynamical systems models directly to human brain imaging data. The MINDy framework is able to produce these data-driven network models for hundreds to thousands of interacting brain regions in just 1-3 minutes per subject. We demonstrate that the models are valid, reliable, and robust. We show that MINDy models are predictive of individualized patterns of resting-state brain dynamical activity. Furthermore, MINDy is better able to uncover the mechanisms underlying individual differences in resting state activity than functional connectivity methods.

Download Full-text

A spatiotemporal complexity architecture of human brain activity

10.1101/2021.06.04.446948 ◽

2021 ◽

Author(s):

Stephan Krohn ◽

Nina von Schwanenflug ◽

Leonhard Waschke ◽

Amy Romanello ◽

Martin Gell ◽

...

Keyword(s):

Human Brain ◽

Neural Activity ◽

Large Scale ◽

Temporal Dynamics ◽

Brain Activity ◽

Brain Regions ◽

Brain Organization ◽

Functional Connections ◽

Signal Complexity ◽

The Brain

The human brain operates in large-scale functional networks, collectively subsumed as the functional connectome1-13. Recent work has begun to unravel the organization of the connectome, including the temporal dynamics of brain states14-20, the trade-off between segregation and integration9,15,21-23, and a functional hierarchy from lower-order unimodal to higher-order transmodal processing systems24-27. However, it remains unknown how these network properties are embedded in the brain and if they emerge from a common neural foundation. Here we apply time-resolved estimation of brain signal complexity to uncover a unifying principle of brain organization, linking the connectome to neural variability6,28-31. Using functional magnetic resonance imaging (fMRI), we show that neural activity is marked by spontaneous "complexity drops" that reflect episodes of increased pattern regularity in the brain, and that functional connections among brain regions are an expression of their simultaneous engagement in such episodes. Moreover, these complexity drops ubiquitously propagate along cortical hierarchies, suggesting that the brain intrinsically reiterates its own functional architecture. Globally, neural activity clusters into temporal complexity states that dynamically shape the coupling strength and configuration of the connectome, implementing a continuous re-negotiation between cost-efficient segregation and communication-enhancing integration9,15,21,23. Furthermore, complexity states resolve the recently discovered association between anatomical and functional network hierarchies comprehensively25-27,32. Finally, brain signal complexity is highly sensitive to age and reflects inter-individual differences in cognition and motor function. In sum, we identify a spatiotemporal complexity architecture of neural activity — a functional "complexome" that gives rise to the network organization of the human brain.

Download Full-text