fMRI Datasets Research Articles

Disruptions of Hierarchical Cortical Organization in Early Psychosis and Schizophrenia

BackgroundThe cerebral cortex is organized hierarchically along an axis that spans unimodal sensorimotor to transmodal association areas. This hierarchy is often characterized using low-dimensional embeddings, termed gradients, of interregional functional coupling estimates measured with resting-state functional magnetic resonance imaging. Such analyses may offer insights into the pathophysiology of schizophrenia, which has been frequently linked to dysfunctional interactions between association and sensorimotor areas. MethodsTo examine disruptions of hierarchical cortical function across distinct stages of psychosis, we applied diffusion map embedding to 2 independent functional magnetic resonance imaging datasets: one comprising 114 patients with early psychosis and 48 control participants, and the other comprising 50 patients with established schizophrenia and 121 control participants. Then, we analyzed the primary sensorimotor-to-association and secondary visual-to-sensorimotor gradients of each participant in both datasets. ResultsThere were no significant differences in regional gradient scores between patients with early psychosis and control participants. Patients with established schizophrenia showed significant differences in the secondary, but not primary, gradient compared with control participants. Gradient differences in schizophrenia were characterized by lower within-network dispersion in the dorsal attention (false discovery rate [FDR]–corrected p [pFDR] < .001), visual (pFDR = .003), frontoparietal (pFDR = .018), and limbic (pFDR = .020) networks and lower between-network dispersion between the visual network and other networks (pFDR < .001). ConclusionsThese findings indicate that differences in cortical hierarchical function occur along the secondary visual-to-sensorimotor axis rather than the primary sensorimotor-to-association axis as previously thought. The absence of differences in early psychosis suggests that visual-sensorimotor abnormalities may emerge as the illness progresses.

Unsupervised contrastive graph learning for resting-state functional MRI analysis and brain disorder detection.

Resting-state functional magnetic resonance imaging (rs-fMRI) helps characterize regional interactions that occur in the human brain at a resting state. Existing research often attempts to explore fMRI biomarkers that best predict brain disease progression using machine/deep learning techniques. Previous fMRI studies have shown that learning-based methods usually require a large amount of labeled training data, limiting their utility in clinical practice where annotating data is often time-consuming and labor-intensive. To this end, we propose an unsupervised contrastive graph learning (UCGL) framework for fMRI-based brain disease analysis, in which a pretext model is designed to generate informative fMRI representations using unlabeled training data, followed by model fine-tuning to perform downstream disease identification tasks. Specifically, in the pretext model, we first design a bi-level fMRI augmentation strategy to increase the sample size by augmenting blood-oxygen-level-dependent (BOLD) signals, and then employ two parallel graph convolutional networks for fMRI feature extraction in an unsupervised contrastive learning manner. This pretext model can be optimized on large-scale fMRI datasets, without requiring labeled training data. This model is further fine-tuned on to-be-analyzed fMRI data for downstream disease detection in a task-oriented learning manner. We evaluate the proposed method on three rs-fMRI datasets for cross-site and cross-dataset learning tasks. Experimental results suggest that the UCGL outperforms several state-of-the-art approaches in automated diagnosis of three brain diseases (i.e., major depressive disorder, autism spectrum disorder, and Alzheimer's disease) with rs-fMRI data.

Invited Session III: Neural network models of the visual system: Exploring the function of different forms of visual recurrence with artificial neural networks.

Behavioral studies suggest that recurrence in the visual system is important for processing degraded stimuli. There are two broad anatomical forms this recurrence can take, lateral or feedback, each with different assumed functions. I'll discuss work wherein I add four different kinds of recurrence-two of each anatomical form-to a feedforward convolutional neural network and find all forms capable of increasing the ability of the network to classify noisy digit images. By using several analysis tools frequently applied to neural data, the distinct strategies used by different networks were identified. The analyses used here can be applied to real neural recordings to identify the strategies at play in the brain. An analysis of an fMRI dataset weakly supports the predictive feedback model but points to a need for higher-resolution cross-regional data to understand recurrent visual processing.

Invited Session III: Neural network models of the visual system: Neural population geometry: An approach for understanding biological and artificial neural networks.

A central goal in neuroscience is to understand how orchestrated computations in the brain arise from the properties of single neurons and networks of such neurons. Answering this question requires theoretical advances that shine light into the 'black box' of representations in neural circuits. In this talk, we will demonstrate theoretical approaches that help describe how cognitive and behavioral task implementations emerge from the structure in neural populations and from biologically plausible neural networks. First, we will introduce an analytic theory that connects geometric structures that arise from neural responses (i.e., neural manifolds) to the neural population's efficiency in implementing a task. In particular, this theory describes a perceptron's capacity for linearly classifying object categories based on the underlying neural manifolds' structural properties. Next, we will describe how such methods can, in fact, open the 'black box' of distributed neuronal circuits in a range of experimental neural datasets. In particular, our method overcomes the limitations of traditional dimensionality reduction techniques, as it operates directly on the high-dimensional representations, rather than relying on low-dimensionality assumptions for visualization. Furthermore, this method allows for simultaneous multi-level analysis, by measuring geometric properties in neural population data, and estimating the amount of task information embedded in the same population. These geometric frameworks are general and can be used across different brain areas and task modalities, as demonstrated in the work of ours and others, ranging from the visual cortex to parietal cortex to hippocampus, and from calcium imaging to electrophysiology to fMRI datasets. Finally, we will discuss our recent efforts to fully extend this multi-level description of neural populations, by (1) investigating how single neuron properties shape the representation geometry in early sensory areas, and by (2) understanding how task-efficient neural manifolds emerge in biologically-constrained neural networks. By extending our mathematical toolkit for analyzing representations underlying complex neuronal networks, we hope to contribute to the long-term challenge of understanding the neuronal basis of tasks and behaviors.

Functional brain network identification and fMRI augmentation using a VAE-GAN framework

Recently, deep learning models have achieved superior performance for mapping functional brain networks from functional magnetic resonance imaging (fMRI) data compared with traditional methods. However, due to the lack of sufficient data and the high dimensionality of brain volume, deep learning models of fMRI tend to suffer from overfitting. In addition, existing methods rarely studied fMRI data augmentation and its application. To address these issues, we developed a VAE-GAN framework that combined a VAE (variational auto-encoder) with a GAN (generative adversarial net) for functional brain network identification and fMRI augmentation. As a generative model, the VAE-GAN models the distribution of fMRI so that it enables the extraction of more generalized features, and thus relieve the overfitting issue. The VAE-GAN is easier to train on fMRI than a standard GAN since it uses latent variables from VAE to generate fake data rather than relying on random noise that is used in a GAN, and it can generate higher quality of fake data than VAE since the discriminator can promote the training of the generator. In other words, the VAE-GAN inherits the advantages of VAE and GAN and avoids their limitations in modeling of fMRI data. Extensive experiments on task fMRI datasets from HCP have proved the effectiveness and superiority of the proposed VAE-GAN framework for identifying both temporal features and functional brain networks compared with existing models, and the quality of fake data is higher than those from VAE and GAN. The results on resting state fMRI of Attention Deficit Hyperactivity Disorder (ADHD)-200 dataset further demonstrated that the fake data generated by the VAE-GAN can help improve the performance of brain network modeling and ADHD classification.

Denoising brain networks using a fixed mathematical phase change in independent component analysis of magnitude-only fMRI data.

Brain networks extracted by independent component analysis (ICA) from magnitude-only fMRI data are usually denoised using various amplitude-based thresholds. By contrast, spatial source phase (SSP) or the phase information of ICA brain networks extracted from complex-valued fMRI data, has provided a simple yet effective way to perform the denoising using a fixed phase change. In this work, we extend the approach to magnitude-only fMRI data to avoid testing various amplitude thresholds for denoising magnitude maps extracted by ICA, as most studies do not save the complex-valued data. The main idea is to generate a mathematical SSP map for a magnitude map using a mapping framework, and the mapping framework is built using complex-valued fMRI data with a known SSP map. Here we leverage the fact that the phase map derived from phase fMRI data has similar phase information to the SSP map. After verifying the use of the magnitude data of complex-valued fMRI, this framework is generalized to work with magnitude-only data, allowing use of our approach even without the availability of the corresponding phase fMRI datasets. We test the proposed method using both simulated and experimental fMRI data including complex-valued data from University of New Mexico and magnitude-only data from Human Connectome Project. The results provide evidence that the mathematical SSP denoising with a fixed phase change is effective for denoising spatial maps from magnitude-only fMRI data in terms of retaining more BOLD-related activity and fewer unwanted voxels, compared with amplitude-based thresholding. The proposed method provides a unified and efficient SSP approach to denoise ICA brain networks in fMRI data.

A large-scale fMRI dataset for the visual processing of naturalistic scenes

One ultimate goal of visual neuroscience is to understand how the brain processes visual stimuli encountered in the natural environment. Achieving this goal requires records of brain responses under massive amounts of naturalistic stimuli. Although the scientific community has put a lot of effort into collecting large-scale functional magnetic resonance imaging (fMRI) data under naturalistic stimuli, more naturalistic fMRI datasets are still urgently needed. We present here the Natural Object Dataset (NOD), a large-scale fMRI dataset containing responses to 57,120 naturalistic images from 30 participants. NOD strives for a balance between sampling variation between individuals and sampling variation between stimuli. This enables NOD to be utilized not only for determining whether an observation is generalizable across many individuals, but also for testing whether a response pattern is generalized to a variety of naturalistic stimuli. We anticipate that the NOD together with existing naturalistic neuroimaging datasets will serve as a new impetus for our understanding of the visual processing of naturalistic stimuli.

Intersubject similarity in neural representations underlies shared episodic memory content

Individuals generally form their unique memories from shared experiences, yet the neural representational mechanisms underlying this subjectiveness of memory are poorly understood. The current study addressed this important question from the cross-subject neural representational perspective, leveraging a large functional magnetic resonance imaging dataset (n = 415) of a face-name associative memory task. We found that individuals' memory abilities were predicted by their synchronization to the group-averaged, canonical trial-by-trial activation level and, to a lesser degree, by their similarity to the group-averaged representational patterns during encoding. More importantly, the memory content shared between pairs of participants could be predicted by their shared local neural activation pattern, particularly in the angular gyrus and ventromedial prefrontal cortex, even after controlling for differences in memory abilities. These results uncover neural representational mechanisms for individualized memory and underscore the constructive nature of episodic memory.

Correntropy-Based Logistic Regression with Automatic Relevance Determination for Robust Sparse Brain Activity Decoding.

Recent studies have used sparse classifications to predict categorical variables from high-dimensional brain activity signals to expose human's mental states and intentions, selecting the relevant features automatically in the model training process. However, existing sparse classification models will likely be prone to the performance degradation which is caused by the noise inherent in the brain recordings. To address this issue, we aim to propose a new robust and sparse classification algorithm in this study. To this end, we introduce the correntropy learning framework into the automatic relevance determination based sparse classification model, proposing a new correntropy-based robust sparse logistic regression algorithm. To demonstrate the superior brain activity decoding performance of the proposed algorithm, we evaluate it on a synthetic dataset, an electroencephalogram (EEG) dataset, and a functional magnetic resonance imaging (fMRI) dataset. The extensive experimental results confirm that not only the proposed method can achieve higher classification accuracy in a noisy and high-dimensional classification task, but also it would select those more informative features for the decoding tasks. Integrating the correntropy learning approach with the automatic relevance determination technique will significantly improve the robustness with respect to the noise, leading to more adequate robust sparse brain decoding algorithm. It provides a more powerful approach in the real-world brain activity decoding and the brain-computer interfaces.

Cneuromod-things : a large-scale fMRI dataset for task- and data-driven assessment of object representation and visual memory recognition in the human brain

cneuromod-things : a large-scale fMRI dataset for task- and data-driven assessment of object representation and visual memory recognition in the human brain

Mitigating susceptibility-induced distortions in high-resolution 3DEPI fMRI at 7T

Geometric distortion is a major limiting factor for spatial specificity in high-resolution fMRI using EPI readouts and is exacerbated at higher field strengths due to increased B0 field inhomogeneity. Prominent correction schemes are based on B0 field-mapping or acquiring reverse phase-encoded (reversed-PE) data. However, to date, comparisons of these techniques in the context of fMRI have only been performed on 2DEPI data, either at lower field or lower resolution. In this study, we investigate distortion compensation in the context of sub-millimetre 3DEPI data at 7T. B0 field-mapping and reversed-PE distortion correction techniques were applied to both partial coverage BOLD-weighted and whole brain MT-weighted 3DEPI data with matched distortion. Qualitative assessment showed overall improvement in cortical alignment for both correction techniques in both 3DEPI fMRI and whole-brain MT-3DEPI datasets. The distortion-corrected MT-3DEPI images were quantitatively evaluated by comparing cortical alignment with an anatomical reference using dice coefficient (DC) and correlation ratio (CR) measures. These showed that B0 field-mapping and reversed-PE methods both improved correspondence between the MT-3DEPI and anatomical data, with more substantial improvements consistently obtained using the reversed-PE approach. Regional analyses demonstrated that the largest benefit of distortion correction, and in particular of the reversed-PE approach, occurred in frontal and temporal regions where susceptibility-induced distortions are known to be greatest, but had not led to complete signal dropout. In conclusion, distortion correction based on reversed-PE data has shown the greater capacity for achieving faithful alignment with anatomical data in the context of high-resolution fMRI at 7T using 3DEPI.

Evaluating deep transfer learning for whole-brain cognitive decoding

Research in many fields has shown that transfer learning (TL) is well-suited to improve the performance of deep learning (DL) models in datasets with small numbers of samples. This empirical success has triggered interest in the application of TL to cognitive decoding analyses with functional neuroimaging data. Here, we systematically evaluate TL for the application of DL models to the decoding of cognitive states (e.g., viewing images of faces or houses) from whole-brain functional Magnetic Resonance Imaging (fMRI) data. We first pre-train two DL architectures on a large, public fMRI dataset and subsequently evaluate their performance in an independent experimental task and a fully independent dataset. The pre-trained DL models consistently achieve higher decoding accuracies and generally require less training time and data than model variants that were not pre-trained, while also outperforming linear baseline models trained from scratch, clearly underlining the benefits of pre-training. We demonstrate that these benefits arise from the ability of the pre-trained models to reuse many of their learned features when training with new data, providing deeper insights into the mechanisms giving rise to the benefits of pre-training. Yet, we also surface nuanced challenges for whole-brain cognitive decoding with DL models when interpreting the decoding decisions of the pre-trained models, as these have learned to utilize the fMRI data in unforeseen and counterintuitive ways to identify individual cognitive states.

A large-scale fMRI dataset for human action recognition

Human action recognition is a critical capability for our survival, allowing us to interact easily with the environment and others in everyday life. Although the neural basis of action recognition has been widely studied using a few action categories from simple contexts as stimuli, how the human brain recognizes diverse human actions in real-world environments still needs to be explored. Here, we present the Human Action Dataset (HAD), a large-scale functional magnetic resonance imaging (fMRI) dataset for human action recognition. HAD contains fMRI responses to 21,600 video clips from 30 participants. The video clips encompass 180 human action categories and offer a comprehensive coverage of complex activities in daily life. We demonstrate that the data are reliable within and across participants and, notably, capture rich representation information of the observed human actions. This extensive dataset, with its vast number of action categories and exemplars, has the potential to deepen our understanding of human action recognition in natural environments.

Novel Linear and Nonlinear Features for the Analysis of Dynamic Brain Functional Connectivity

Dynamic functional connectivity (dFC), estimating changes in the brain interaction, has been employed as a sensitive biomarker for identifying neurological and psychological disorders. Deep learning is frequently used method for the analysis of dFC, but it is unfortunately unexplainable and computational expensive. Root mean square (RMS) value is also proposed but may be insufficient for a complete characterization of the dFC. The present study aims at investigating the feasibility of several novel features for dFC analysis. Two public datasets were used, i.e., a rest-state functional magnetic resonance imaging (fMRI) dataset containing healthy controls (HC) and Alzheimer’s Disease (AD) patients and an electroencephalography (EEG) dataset containing two motor imagery tasks (cylindrical and lumbrical grasping). The dFC was estimated using the Pearson’s correlation with a sliding window. Ten features were extracted from pairwise Pearson’s correlation of the dFC. A Wilcoxon rank sum test was applied to each feature calculated from the two classes of each dataset. The area under the receiver operating characteristic curve (AUC) was calculated to assess the discrimination power of each feature. Our results show that 6014 out of 66700 features are significantly different between HC and AD, and 6670 out of 17700 features between cylindrical and lumbrical grasping. The maximum AUCs for individual features are 0.69 and 0.65 for the fMRI and EEG datasets, respectively. These results confirm that, in addition to RMS, other categories of features contribute also significantly to the analysis of dFC.

BOLD Response is more than just magnitude: Improving detection sensitivity through capturing hemodynamic profiles

Typical fMRI analyses often assume a canonical hemodynamic response function (HRF) that primarily focuses on the peak height of the overshoot, neglecting other morphological aspects. Consequently, reported analyses often reduce the overall response curve to a single scalar value. In this study, we take a data-driven approach to HRF estimation at the whole-brain voxel level, without assuming a response profile at the individual level. We then employ a roughness penalty at the population level to estimate the response curve, aiming to enhance predictive accuracy, inferential efficiency, and cross-study reproducibility. By examining a fast event-related FMRI dataset, we demonstrate the shortcomings and information loss associated with adopting the canonical approach. Furthermore, we address the following key questions: 1) To what extent does the HRF shape vary across different regions, conditions, and participant groups? 2) Does the data-driven approach improve detection sensitivity compared to the canonical approach? 3) Can analyzing the HRF shape help validate the presence of an effect in conjunction with statistical evidence? 4) Does analyzing the HRF shape offer evidence for whole-brain response during a simple task?

Random matrix theory tools for the predictive analysis of functional magnetic resonance imaging examinations.

Random matrix theory (RMT) is an increasingly useful tool for understanding large, complex systems. Prior studies have examined functional magnetic resonance imaging (fMRI) scans using tools from RMT, with some success. However, RMT computations are highly sensitive to a number of analytic choices, and the robustness of findings involving RMT remains in question. We systematically investigate the usefulness of RMT on a wide variety of fMRI datasets using a rigorous predictive framework. We develop open-source software to efficiently compute RMT features from fMRI images and examine the cross-validated predictive potential of eigenvalue and RMT-based features ("eigenfeatures") with classic machine-learning classifiers. We systematically vary pre-processing extent, normalization procedures, RMT unfolding procedures, and feature selection and compare the impact of these analytic choices on the distributions of cross-validated prediction performance for each combination of dataset binary classification task, classifier, and feature. To deal with class imbalance, we use the area under the receiver operating characteristic curve (AUROC) as the main performance metric. Across all classification tasks and analytic choices, we find RMT- and eigenvalue-based "eigenfeatures" to have predictive utility more often than not (82.4% of median ; median AUROC range across classification tasks 0.47 to 0.64). Simple baseline reductions on source timeseries, by contrast, were less useful (58.8% of median , median AUROC range across classification tasks 0.42 to 0.62). Additionally, eigenfeature AUROC distributions were overall more right-tailed than baseline features, suggesting greater predictive potential. However, performance distributions were wide and often significantly affected by analytic choices. Eigenfeatures clearly have potential for understanding fMRI functional connectivity in a wide variety of scenarios. The utility of these features is strongly dependent on analytic decisions, suggesting caution when interpreting past and future studies applying RMT to fMRI. However, our study demonstrates that the inclusion of RMT statistics in fMRI investigations could improve prediction performances across a wide variety of phenomena.

Smoothness Harmonic: A Graph-Based Approach to Reveal Spatiotemporal Patterns of Cortical Dynamics in fMRI Data

Despite fMRI data being interpreted as time-varying graphs in graph analysis, there has been more emphasis on learning sophisticated node embeddings and complex graph structures rather than providing a macroscopic description of cortical dynamics. In this paper, I introduce the notion of smoothness harmonics to capture the slowly varying cortical dynamics in graph-based fMRI data in the form of spatiotemporal smoothness patterns. These smoothness harmonics are rooted in the eigendecomposition of graph Laplacians, which reveal how low-frequency-dominated fMRI signals propagate across the cortex and through time. We showcase their usage in a real fMRI dataset to differentiate the cortical dynamics of children and adults while also demonstrating their empirical merit over the static functional connectomes in inter-subject and between-group classification analyses.

A Scalable Approach to Independent Vector Analysis by Shared Subspace Separation for Multi-Subject fMRI Analysis.

Joint blind source separation (JBSS) has wide applications in modeling latent structures across multiple related datasets. However, JBSS is computationally prohibitive with high-dimensional data, limiting the number of datasets that can be included in a tractable analysis. Furthermore, JBSS may not be effective if the data's true latent dimensionality is not adequately modeled, where severe overparameterization may lead to poor separation and time performance. In this paper, we propose a scalable JBSS method by modeling and separating the "shared" subspace from the data. The shared subspace is defined as the subset of latent sources that exists across all datasets, represented by groups of sources that collectively form a low-rank structure. Our method first provides the efficient initialization of the independent vector analysis (IVA) with a multivariate Gaussian source prior (IVA-G) specifically designed to estimate the shared sources. Estimated sources are then evaluated regarding whether they are shared, upon which further JBSS is applied separately to the shared and non-shared sources. This provides an effective means to reduce the dimensionality of the problem, improving analyses with larger numbers of datasets. We apply our method to resting-state fMRI datasets, demonstrating that our method can achieve an excellent estimation performance with significantly reduced computational costs.

Classification of Cognitive States using Task-Specific Connectivity Features

Human brain activity maps are produced by functional MRI (fMRI) research that describes the average level of engagement during a specific task of various brain regions. Functional connectivity describes the interrelationship, integrated performance, and organization of these different brain regions. This study investigates functional connectivity to quantify the interactions between different brain regions engaged concurrently in a specific task. The key focus of this study was to introduce and demonstrate task-specific functional connectivity among brain regions using fMRI data and decode cognitive states by proposing a novel classifier using connectivity features. Two connectivity models were considered: a graph-based task-specific functional connectivity and a Granger causality-transfer entropy framework. Connectivity strengths obtained among brain regions were used for cognitive state classification. The parameters of the nodal and global graph analysis from the graph-based connectivity framework were considered, and the transfer entropy values of the causal connectivity model were considered as features for the cognitive state classification. The proposed model achieved an average accuracy of 95% on the StarPlus fMRI dataset and showed an improvement of 5% compared to the existing Tensor-SVD classification algorithm.

Visual Image Decoding of Brain Activities Using a Dual Attention Hierarchical Latent Generative Network With Multiscale Feature Fusion

Reconstructing visual stimulus from human brain activity measured with functional magnetic resonance imaging (fMRI) is a challenging decoding task for revealing the visual system. Recent deep learning approaches commonly neglect the relationship between hierarchical image features and different regions of the visual cortex, and fail to use global and local image features in reconstructing visual stimulus. To address these issues, in this paper, a novel neural decoding framework is proposed by using a dual attention hierarchical latent generative network with multi-scale feature fusion (DA-HLGN-MSFF) method. Specifically, the fMRI data is firstly encoded to hierarchical features of our image encoder network which employs a multi-kernel convolution block to extract the multi-scale spatial information of images. In order to reconstruct the perceived images and further improve the performance of our generator network, a dual attention block based on channel-spatial attention mechanism is then proposed to exploit the inter-channel relationships and spatial long-range dependencies of features. Moreover, a multi-scale feature fusion block is finally adopted to aggregate the global and local information of features at different scales and synthesize the final reconstructed images in the generator network. Competitive experimental results on two public fMRI datasets demonstrate that our method is able to achieve promising reconstructing performance compared with the state-of-the-art methods. The codes of our proposed DA-HLGN-MSFF method will be open access on https://github.com/ljbuaa/HLDAGN.