Search

Non-negligible idiosyncrasy due to interindividual differences is an ongoing issue in resting-state functional MRI (rfMRI) analysis. We show that a deep neural network (DNN) can be employed for individual identification by learning important features from the time-varying functional connectivity (FC) of rfMRI in the Human Connectome Project. We employed the trained DNN to identify individuals from an independent dataset acquired at our institution. The results revealed that the DNN could successfully identify 300 individuals with an error rate of 2.9% using 15 s time-window and 870 individuals with an error rate of 6.7%. A trained DNN with nonlinear hidden layers led to the proposal of the "fingerprint of FC" (fpFC) as representative edges of individual FC. The fpFCs for individuals exhibited commonly important and individual-specific edges across time-window lengths (from 5 min to 15 s). Furthermore, the utility of our model for another group of subjects was validated, supporting the feasibility of our technique in the context of transfer learning. In conclusion, our study offers an insight into the discovery of the intrinsic mode of the human brain using whole-brain resting-state FC and DNNs., (© 2023 The Authors. Human Brain Mapping published by Wiley Periodicals LLC.)

Objective: Functional connectivity in intrinsic brain networks, namely, the triple network, which includes the salience network, default mode network (DMN) and central executive network (CEN), has been suggested as prominent, major networks involved in human cognition and mental state-mindfulness, mind-wandering and resting-state. Despite the established roles of functional connections within and between intrinsic networks, there has been limited research on the effective connectivity of mindfulness, mind-wandering and resting-state using the triple network, as well as on their direct comparisons., Methods: We employed spectral dynamic causal modeling to compare effective connectivity patterns across mindfulness (i.e. attention focused on physical sensations of breathing), mind-wandering (i.e. connecting thoughts) and resting-state (i.e. relaxing while remaining calm and awake) conditions using functional MRI data of healthy subjects who underwent ambulatory training by practicing mindfulness and mind-wandering (N = 59)., Results: When comparing mindfulness and mindwandering conditions, our analysis results revealed that salience network and CEN interacted depending on mindfulness or mind-wandering. When mindfulness or mind-wandering was compared to resting-state, mindfulness increased the effective connectivity from the left CEN to salience network through DMN, whereas mindwandering increased the effective connectivity from the DMN to right CEN., Conclusion: To the best of our knowledge, this is the first study to examine possible differences in effective connectivity patterns among mindfulness, mind-wandering and resting-state using the triple network. We believe that our findings will provide deeper insights into the neural substrates of mindfulness compared to mind-wandering and resting-state., (Copyright © 2022 The Author(s). Published by Wolters Kluwer Health, Inc.)

Transcranial focused ultrasound (tFUS) is a promising non-invasive technique for treating neurological and psychiatric disorders. One of the challenges for tFUS is the disruption of wave propagation through the skull. Consequently, despite the risks associated with exposure to ionizing radiation, computed tomography (CT) is required to estimate the acoustic transmission through the skull. This study aims to generate synthetic CT (sCT) from T1-weighted magnetic resonance imaging (MRI) and investigate its applicability to tFUS acoustic simulation. We trained a 3D conditional generative adversarial network (3D-cGAN) with 15 subjects. We then assessed image quality with 15 test subjects: mean absolute error (MAE) = 85.72±9.50 HU (head) and 280.25±24.02 HU (skull), dice coefficient similarity (DSC) = 0.88±0.02 (skull). In terms of skull density ratio (SDR) and skull thickness (ST), no significant difference was found between sCT and real CT (rCT). When the acoustic simulation results of rCT and sCT were compared, the intracranial peak acoustic pressure ratio was found to be less than 4%, and the distance between focal points less than 1 mm.

Real-time fMRI neurofeedback is an increasingly popular neuroimaging technique that allows an individual to gain control over his/her own brain signals, which can lead to improvements in behavior in healthy participants as well as to improvements of clinical symptoms in patient populations. However, a considerably large ratio of participants undergoing neurofeedback training do not learn to control their own brain signals and, consequently, do not benefit from neurofeedback interventions, which limits clinical efficacy of neurofeedback interventions. As neurofeedback success varies between studies and participants, it is important to identify factors that might influence neurofeedback success. Here, for the first time, we employed a big data machine learning approach to investigate the influence of 20 different design-specific (e.g. activity vs. connectivity feedback), region of interest-specific (e.g. cortical vs. subcortical) and subject-specific factors (e.g. age) on neurofeedback performance and improvement in 608 participants from 28 independent experiments. With a classification accuracy of 60% (considerably different from chance level), we identified two factors that significantly influenced neurofeedback performance: Both the inclusion of a pre-training no-feedback run before neurofeedback training and neurofeedback training of patients as compared to healthy participants were associated with better neurofeedback performance. The positive effect of pre-training no-feedback runs on neurofeedback performance might be due to the familiarization of participants with the neurofeedback setup and the mental imagery task before neurofeedback training runs. Better performance of patients as compared to healthy participants might be driven by higher motivation of patients, higher ranges for the regulation of dysfunctional brain signals, or a more extensive piloting of clinical experimental paradigms. Due to the large heterogeneity of our dataset, these findings likely generalize across neurofeedback studies, thus providing guidance for designing more efficient neurofeedback studies specifically for improving clinical neurofeedback-based interventions. To facilitate the development of data-driven recommendations for specific design details and subpopulations the field would benefit from stronger engagement in open science research practices and data sharing., Competing Interests: Declaration of Competing Interest SHK receives payment from Mendi Innovations AB, Stockholm, Sweden., (Copyright © 2021 The Author(s). Published by Elsevier Inc. All rights reserved.)

Deep-learning methods based on deep neural networks (DNNs) have recently been successfully utilized in the analysis of neuroimaging data. A convolutional neural network (CNN) is a type of DNN that employs a convolution kernel that covers a local area of the input sample and moves across the sample to provide a feature map for the subsequent layers. In our study, we hypothesized that a 3D-CNN model with down-sampling operations such as pooling and/or stride would have the ability to extract robust feature maps from the shifted and scaled neuronal activations in a single functional MRI (fMRI) volume for the classification of task information associated with that volume. Thus, the 3D-CNN model would be able to ameliorate the potential misalignment of neuronal activations and over-/under-activation in local brain regions caused by imperfections in spatial alignment algorithms, confounded by variability in blood-oxygenation-level-dependent (BOLD) responses across sessions and/or subjects. To this end, the fMRI volumes acquired from four sensorimotor tasks (left-hand clenching, right-hand clenching, auditory attention, and visual stimulation) were used as input for our 3D-CNN model to classify task information using a single fMRI volume. The classification performance of the 3D-CNN was systematically evaluated using fMRI volumes obtained from various minimal preprocessing scenarios applied to raw fMRI volumes that excluded spatial normalization to a template and those obtained from full preprocessing that included spatial normalization. Alternative classifier models such as the 1D fully connected DNN (1D-fcDNN) and support vector machine (SVM) were also used for comparison. The classification performance was also assessed for several k-fold cross-validation (CV) schemes, including leave-one-subject-out CV (LOOCV). Overall, the classification results of the 3D-CNN model were superior to that of the 1D-fcDNN and SVM models. When using the fully-processed fMRI volumes with LOOCV, the mean error rates (± the standard error of the mean) for the 3D-CNN, 1D-fcDNN, and SVM models were 2.1% (± 0.9), 3.1% (± 1.2), and 4.1% (± 1.5), respectively (p = 0.041 from a one-way ANOVA). The error rates for 3-fold CV were higher (2.4% ± 1.0, 4.2% ± 1.3, and 10.1% ± 2.0; p < 0.0003 from a one-way ANOVA). The mean error rates also increased considerably using the raw fMRI 3D volume data without preprocessing (26.2% for the 3D-CNN, 75.0% for the 1D-fcDNN, and 75.0% for the SVM). Furthermore, the ability of the pre-trained 3D-CNN model to handle shifted and scaled neuronal activations was demonstrated in an online scenario for five-class classification (i.e., four sensorimotor tasks and the resting state) using the real-time fMRI of three participants. The resulting classification accuracy was 78.5% (± 1.4), 26.7% (± 5.9), and 21.5% (± 3.1) for the 3D-CNN, 1D-fcDNN, and SVM models, respectively. The superior performance of the 3D-CNN compared to the 1D-fcDNN was verified by analyzing the resulting feature maps and convolution filters that handled the shifted and scaled neuronal activations and by utilizing an independent public dataset from the Human Connectome Project., Competing Interests: Declaration of Competing Interest The authors have no conflicts of interest regarding this study, including financial, consultant, institutional, or other relationships. The sponsors had no involvement in the study design, data collection, analysis or interpretation of the data, manuscript preparation, or the decision to submit for publication., (Copyright © 2020. Published by Elsevier Inc.)

Competition and collaboration are strategies that can be used to optimize the outcomes of social interactions. Research into the neuronal substrates underlying these aspects of social behavior has been limited due to the difficulty in distinguishing complex activation via univariate analysis. Therefore, we employed multivoxel pattern analysis of functional magnetic resonance imaging to reveal the neuronal activations underlying competitive and collaborative processes when the collaborator/opponent used myopic/predictive reasoning. Twenty-four healthy subjects participated in 2 × 2 matrix-based sequential-move games. Searchlight-based multivoxel patterns were used as input for a support vector machine using nested cross-validation to distinguish game conditions, and identified voxels were validated via the regression of the behavioral data with bootstrapping. The left anterior insula (accuracy = 78.5%) was associated with competition, and middle frontal gyrus (75.1%) was associated with predictive reasoning. The inferior/superior parietal lobules (84.8%) and middle frontal gyrus (84.7%) were associated with competition, particularly in trials with a predictive opponent. The visual/motor areas were related to response time as a proxy for visual attention and task difficulty. Our results suggest that multivoxel patterns better represent the neuronal substrates underlying the social cognition of collaboration and competition intermixed with myopic and predictive reasoning than do univariate features., (© 2020 The Authors. Human Brain Mapping published by Wiley Periodicals LLC.)

Neurofeedback training has been shown to influence behavior in healthy participants as well as to alleviate clinical symptoms in neurological, psychosomatic, and psychiatric patient populations. However, many real-time fMRI neurofeedback studies report large inter-individual differences in learning success. The factors that cause this vast variability between participants remain unknown and their identification could enhance treatment success. Thus, here we employed a meta-analytic approach including data from 24 different neurofeedback studies with a total of 401 participants, including 140 patients, to determine whether levels of activity in target brain regions during pretraining functional localizer or no-feedback runs (i.e., self-regulation in the absence of neurofeedback) could predict neurofeedback learning success. We observed a slightly positive correlation between pretraining activity levels during a functional localizer run and neurofeedback learning success, but we were not able to identify common brain-based success predictors across our diverse cohort of studies. Therefore, advances need to be made in finding robust models and measures of general neurofeedback learning, and in increasing the current study database to allow for investigating further factors that might influence neurofeedback learning., (© 2020 The Authors. Human Brain Mapping published by Wiley Periodicals LLC.)

The naturalistic viewing of a video clip enables participants to obtain more information from the clip compared to conventional viewing of a static image. Because changing the field-of-view (FoV) allows new visual information to be obtained, we were motivated to investigate whether naturalistic viewing with varying FoV based on active eye movement can enhance the viewing experience of natural stimuli, such as those found in a video clip with a 360° FoV in an MRI scanner. To this end, we developed a novel naturalistic viewing paradigm based on real-time eye-gaze tracking while participants were watching a 360° panoramic video during fMRI acquisition. The gaze position of the participants was recorded using an eye-tracking computer and then transmitted to a stimulus presentation computer via a TCP/IP connection. The identified gaze position was then used to alter the participants' FoV of the video clip in real-time, so the participants could change their FoV to fully explore the 360° video clip (referred to in this paper as active viewing). The gaze position of one participant while watching a video was used to change the FoV of the same video clip for a paired participant (referred to as yoked or passive viewing). Four 360° panoramic videos were used as stimuli, divided into categories based on the brightness level (i.e., bright vs. dark) and location (i.e., nature vs. city). Each of the subjects participated in the active viewing of one of the two nature videos and one of the two city videos and then engaged in the passive viewing of the other video in each category, followed by conventional viewing with a fixed FoV (referred to as fixed viewing) after each of the active or passive viewings. Forty-eight healthy volunteers participated in the study, and data from 42 of these participants were used in the analysis. Representational similarity analysis (RSA) was conducted in a multiple regression framework using representational dissimilarity matrix (RDM) codes to accommodate all of the information regarding neuronal activations from fMRI analysis and the participants' subjective ratings of their viewing experience with the four video clips and with the two contrasting viewing conditions (i.e., "active-fixed" and "passive-fixed"). It was found that the participants' naturalistic viewing experience of the video clips was substantially more immersive with active viewing than with passive and fixed viewing. The RSA using the RDM codes revealed the brain regions associated with the viewing experience, including eye movement and spatial navigation in the superior frontal area (of Brodmann's area 6) and the inferior/superior parietal areas, respectively. Brain regions potentially associated with cognitive and affective processing during the viewing of the video, such as the default-mode networks and insular/Rolandic operculum areas, were also identified. To the best of our knowledge, this is the first study that has used the participants' eye movements to interactively change their FoV for 360° panoramic video clips in real-time. Our method of utilizing the MRI environment can be further extended to other environments such as electroencephalography and behavioral research. It would also be feasible to apply our method to virtual reality and/or augmented reality systems to maximize user experience based on their eye movement., Competing Interests: Declaration of competing interest The authors have no conflicts of interest regarding this study, including financial, consultant, institutional, or other relationships. The sponsors had no involvement in the study design, data collection, analysis or interpretation of the data, manuscript preparation, or the decision to submit for publication., (Copyright © 2020 The Authors. Published by Elsevier Inc. All rights reserved.)

Perceptual learning refers to improved perceptual performance after intensive training and was initially suggested to reflect long-term plasticity in early visual cortex. Recent behavioral and neurophysiological evidence further suggested that the plasticity in brain regions related to decision making could also contribute to the observed training effects. However, how perceptual learning modulates the responses of decision-related regions in the human brain remains largely unknown. In the present study, we combined psychophysics and functional magnetic resonance imaging (fMRI), and adopted a model-based approach to investigate this issue. We trained participants on a motion direction discrimination task and fitted their behavioral data using the linear ballistic accumulator model. The results from model fitting showed that behavioral improvement could be well explained by a specific improvement in sensory information accumulation. A critical model parameter, the drift rate of the information accumulation, was correlated with the fMRI responses derived from three spatial independent components: ventral premotor cortex (PMv), supplementary eye field (SEF), and the fronto-parietal network, including intraparietal sulcus (IPS) and frontal eye field (FEF). In this decision network, we found that the behavioral training effects were accompanied by signal enhancement specific to trained direction in PMv and FEF. Further, we also found direction-specific signal reduction in sensory areas (V3A and MT+), as well as the strengthened effective connectivity from V3A to PMv and from IPS to FEF. These findings provide evidence for the learning-induced decision refinement after perceptual learning and the brain regions that are involved in this process.

Feedforward deep neural networks (DNNs), artificial neural networks with multiple hidden layers, have recently demonstrated a record-breaking performance in multiple areas of applications in computer vision and speech processing. Following the success, DNNs have been applied to neuroimaging modalities including functional/structural magnetic resonance imaging (MRI) and positron-emission tomography data. However, no study has explicitly applied DNNs to 3D whole-brain fMRI volumes and thereby extracted hidden volumetric representations of fMRI that are discriminative for a task performed as the fMRI volume was acquired. Our study applied fully connected feedforward DNN to fMRI volumes collected in four sensorimotor tasks (i.e., left-hand clenching, right-hand clenching, auditory attention, and visual stimulus) undertaken by 12 healthy participants. Using a leave-one-subject-out cross-validation scheme, a restricted Boltzmann machine-based deep belief network was pretrained and used to initialize weights of the DNN. The pretrained DNN was fine-tuned while systematically controlling weight-sparsity levels across hidden layers. Optimal weight-sparsity levels were determined from a minimum validation error rate of fMRI volume classification. Minimum error rates (mean±standard deviation; %) of 6.9 (±3.8) were obtained from the three-layer DNN with the sparsest condition of weights across the three hidden layers. These error rates were even lower than the error rates from the single-layer network (9.4±4.6) and the two-layer network (7.4±4.1). The estimated DNN weights showed spatial patterns that are remarkably task-specific, particularly in the higher layers. The output values of the third hidden layer represented distinct patterns/codes of the 3D whole-brain fMRI volume and encoded the information of the tasks as evaluated from representational similarity analysis. Our reported findings show the ability of the DNN to classify a single fMRI volume based on the extraction of hidden representations of fMRI volumes associated with tasks across multiple hidden layers. Our study may be beneficial to the automatic classification/diagnosis of neuropsychiatric and neurological diseases and prediction of disease severity and recovery in (pre-) clinical settings using fMRI volumes without requiring an estimation of activation patterns or ad hoc statistical evaluation., Competing Interests: The authors have no conflicts of interest regarding this study, including financial, consultant, institutional, or other relationships., (Copyright © 2016 Elsevier Inc. All rights reserved.)

Functional connectivity (FC) patterns obtained from resting-state functional magnetic resonance imaging data are commonly employed to study neuropsychiatric conditions by using pattern classifiers such as the support vector machine (SVM). Meanwhile, a deep neural network (DNN) with multiple hidden layers has shown its ability to systematically extract lower-to-higher level information of image and speech data from lower-to-higher hidden layers, markedly enhancing classification accuracy. The objective of this study was to adopt the DNN for whole-brain resting-state FC pattern classification of schizophrenia (SZ) patients vs. healthy controls (HCs) and identification of aberrant FC patterns associated with SZ. We hypothesized that the lower-to-higher level features learned via the DNN would significantly enhance the classification accuracy, and proposed an adaptive learning algorithm to explicitly control the weight sparsity in each hidden layer via L1-norm regularization. Furthermore, the weights were initialized via stacked autoencoder based pre-training to further improve the classification performance. Classification accuracy was systematically evaluated as a function of (1) the number of hidden layers/nodes, (2) the use of L1-norm regularization, (3) the use of the pre-training, (4) the use of framewise displacement (FD) removal, and (5) the use of anatomical/functional parcellation. Using FC patterns from anatomically parcellated regions without FD removal, an error rate of 14.2% was achieved by employing three hidden layers and 50 hidden nodes with both L1-norm regularization and pre-training, which was substantially lower than the error rate from the SVM (22.3%). Moreover, the trained DNN weights (i.e., the learned features) were found to represent the hierarchical organization of aberrant FC patterns in SZ compared with HC. Specifically, pairs of nodes extracted from the lower hidden layer represented sparse FC patterns implicated in SZ, which was quantified by using kurtosis/modularity measures and features from the higher hidden layer showed holistic/global FC patterns differentiating SZ from HC. Our proposed schemes and reported findings attained by using the DNN classifier and whole-brain FC data suggest that such approaches show improved ability to learn hidden patterns in brain imaging data, which may be useful for developing diagnostic tools for SZ and other neuropsychiatric disorders and identifying associated aberrant FC patterns., (Copyright © 2015 Elsevier Inc. All rights reserved.)

Real-time fMRI (rtfMRI) neurofeedback (NF) facilitates volitional control over brain activity and the modulation of associated mental functions. The NF signals of traditional rtfMRI-NF studies predominantly reflect neuronal activity within ROIs. In this study, we describe a novel rtfMRI-NF approach that includes a functional connectivity (FC) component in the NF signal (FC-added rtfMRI-NF). We estimated the efficacy of the FC-added rtfMRI-NF method by applying it to nicotine-dependent heavy smokers in an effort to reduce cigarette craving. ACC and medial pFC as well as the posterior cingulate cortex and precuneus are associated with cigarette craving and were chosen as ROIs. Fourteen heavy smokers were randomly assigned to receive one of two types of NF: traditional activity-based rtfMRI-NF or FC-added rtfMRI-NF. Participants received rtfMRI-NF training during two separate visits after overnight smoking cessation, and cigarette craving score was assessed. The FC-added rtfMRI-NF resulted in greater neuronal activity and increased FC between the targeted ROIs than the traditional activity-based rtfMRI-NF and resulted in lower craving score. In the FC-added rtfMRI-NF condition, the average of neuronal activity and FC was tightly associated with craving score (Bonferroni-corrected p = .028). However, in the activity-based rtfMRI-NF condition, no association was detected (uncorrected p > .081). Non-rtfMRI data analysis also showed enhanced neuronal activity and FC with FC-added NF than with activity-based NF. These results demonstrate that FC-added rtfMRI-NF facilitates greater volitional control over brain activity and connectivity and greater modulation of mental function than activity-based rtfMRI-NF.

Electroencephalography (EEG) data simultaneously acquired with functional magnetic resonance imaging (fMRI) data are preprocessed to remove gradient artifacts (GAs) and ballistocardiographic artifacts (BCAs). Nonetheless, these data, especially in the gamma frequency range, can be contaminated by residual artifacts produced by mechanical vibrations in the MRI system, in particular the cryogenic pump that compresses and transports the helium that chills the magnet (the helium-pump). However, few options are available for the removal of helium-pump artifacts. In this study, we propose a recursive approach of EEG-segment-based principal component analysis (rsPCA) that enables the removal of these helium-pump artifacts. Using the rsPCA method, feature vectors representing helium-pump artifacts were successfully extracted as eigenvectors, and the reconstructed signals of the feature vectors were subsequently removed. A test using simultaneous EEG-fMRI data acquired from left-hand (LH) and right-hand (RH) clenching tasks performed by volunteers found that the proposed rsPCA method substantially reduced helium-pump artifacts in the EEG data and significantly enhanced task-related gamma band activity levels (p=0.0038 and 0.0363 for LH and RH tasks, respectively) in EEG data that have had GAs and BCAs removed. The spatial patterns of the fMRI data were estimated using a hemodynamic response function (HRF) modeled from the estimated gamma band activity in a general linear model (GLM) framework. Active voxel clusters were identified in the post-/pre-central gyri of motor area, only from the rsPCA method (uncorrected p<0.001 for both LH/RH tasks). In addition, the superior temporal pole areas were consistently observed (uncorrected p<0.001 for the LH task and uncorrected p<0.05 for the RH task) in the spatial patterns of the HRF model for gamma band activity when the task paradigm and movement were also included in the GLM., (Copyright © 2014 Elsevier Inc. All rights reserved.)

Purpose: To investigate lateral geniculate nucleus (LGN) volume of primary open-angle glaucoma (POAG) patients compared with age- and sex-matched controls using ultra-high field 7.0-T magnetic resonance imaging (MRI)., Methods: The study included 18 patients with POAG and 18 age- and sex-matched healthy volunteers. All subjects underwent imaging on a high-resolution 7.0-T MRI system. Bilateral LGNs were identified and manually delineated, and LGN volumes were compared. Peripapillary retinal nerve fiber layer (pRNFL) thickness, optic nerve head parameters (including optic disc size, rim area, and cup-to-disc ratio), and combined thickness of the ganglion cell layer and inner plexus layer (GC-IPL) were measured by Cirrus high-definition optical coherence tomography (OCT). Correlations between OCT parameters and LGN volume were investigated., Results: Mean LGN volumes were significantly smaller in the POAG group than in the control group (right, glaucoma 83.97 mm(3) [SD ± 26.65] versus control 106.12 mm(3) [SD ± 24.32]; left, glaucoma 65.12 mm(3) [SD ± 29.41] versus control 92.70 mm(3) [SD ± 24.42], both P < 0.05). In the POAG group, average GC-IPL thickness was correlated with contralateral LGN volume (right LGN: r = 0.605, P = 0.008; left LGN: r = 0.471, P = 0.049). The correlation for right LGN volume remained significant after correction for multiple comparisons. However, there was no correlation between LGN volume and average pRNFL thickness or optic disc parameters in the POAG group., Conclusions: On high-resolution 7.0-T MRI, LGN volumes in POAG patients are significantly smaller than those of healthy subjects. Furthermore, in patients, LGN volume was found to be significantly correlated with GC-IPL thickness of the contralateral eye., (Copyright 2014 The Association for Research in Vision and Ophthalmology, Inc.)

The identification of mild cognitive impairments (MCI) via either structural magnetic resonance imaging (sMRI) or functional MRI (fMRI) has great potential due to the non-invasiveness of the techniques. Furthermore, these techniques allow longitudinal follow-ups of single subjects via repeated measurements. sMRI- or fMRI-based biomarkers have been adopted separately to diagnose MCI; however, there has not been a systematic effort to integrate sMRI- and fMRI-based features to increase MCI detection accuracy. This study investigated whether the detection of MCI can be improved via the integration of biomarkers identified from both sMRI and fMRI modalities. Regional volume sizes and neuronal activity levels of brains from MCI subjects were compared with those from healthy controls and used to identify biomarkers from sMRI and fMRI data, respectively. In the subsequent classification phase, MCI was automatically detected using a support vector machine algorithm that employed the identified sMRI- and fMRI-based biomarkers as an input feature vector. The results indicate that the fMRI-based biomarkers provided more information for detecting MCI than the sMRI-based biomarkers. Moreover, the integrated feature sets using the sMRI- and fMRI-based biomarkers consistently showed greater detection accuracy than the feature sets based only on the fMRI-based biomarkers. The results demonstrate that integration of sMRI and fMRI modalities can provide supplemental information to improve the diagnosis of MCI relative to either the sMRI or fMRI modalities alone., (Copyright © 2013 Elsevier Inc. All rights reserved.)

Nonnegative matrix factorization (NMF) is a blind source separation (BSS) algorithm which is based on the distinct constraint of nonnegativity of the estimated parameters as well as on the measured data. In this study, according to the potential feasibility of NMF for fMRI data, the four most popular NMF algorithms, corresponding to the following two types of (1) least-squares based update [i.e., alternating least-squares NMF (ALSNMF) and projected gradient descent NMF] and (2) multiplicative update (i.e., NMF based on Euclidean distance and NMF based on divergence cost function), were investigated by using them to estimate task-related neuronal activities. These algorithms were applied firstly to individual data from a single subject and, subsequently, to group data sets from multiple subjects. On the single-subject level, although all four algorithms detected task-related activation from simulated data, the performance of multiplicative update NMFs was significantly deteriorated when evaluated using visuomotor task fMRI data, for which they failed in estimating any task-related neuronal activities. In group-level analysis on both simulated data and real fMRI data, ALSNMF outperformed the other three algorithms. The presented findings may suggest that ALSNMF appears to be the most promising option among the tested NMF algorithms to extract task-related neuronal activities from fMRI data., (Copyright © 2013 Elsevier Inc. All rights reserved.)

We determined the association of neuronal circuitry with the desire to smoke by acquiring and analyzing functional MRI data. The data were acquired in both abstained and subsequently satiated (by 'natural' cigarette smoking) heavy smokers and also in demographically and intellectually matched nonsmokers. During the acquisition, participants were viewing alternating smoking and nonsmoking images that were interleaved by fixation images. From the results, the activities in the mesocorticolimbic pathway including the orbitofrontal cortex, parahippocampus, hippocampus, and midbrain were significantly negatively correlated with carbon monoxide (CO) levels. In contrast, the activities in the motor area and the posterior cingulate cortex plus precuneus were significantly positively correlated with the CO levels. This is the first study to show that mesocorticolimbic and midbrain activities are strongly associated with CO levels, and therefore, possibly with smoking desire levels because of the strong correlation between CO levels and blood nicotine levels.

This study proposes an iterative dual-regression (DR) approach with sparse prior regularization to better estimate an individual's neuronal activation using the results of an independent component analysis (ICA) method applied to a temporally concatenated group of functional magnetic resonance imaging (fMRI) data (i.e., Tc-GICA method). An ordinary DR approach estimates the spatial patterns (SPs) of neuronal activation and corresponding time courses (TCs) specific to each individual's fMRI data with two steps involving least-squares (LS) solutions. Our proposed approach employs iterative LS solutions to refine both the individual SPs and TCs with an additional a priori assumption of sparseness in the SPs (i.e., minimally overlapping SPs) based on L(1)-norm minimization. To quantitatively evaluate the performance of this approach, semi-artificial fMRI data were created from resting-state fMRI data with the following considerations: (1) an artificially designed spatial layout of neuronal activation patterns with varying overlap sizes across subjects and (2) a BOLD time series (TS) with variable parameters such as onset time, duration, and maximum BOLD levels. To systematically control the spatial layout variability of neuronal activation patterns across the "subjects" (n=12), the degree of spatial overlap across all subjects was varied from a minimum of 1 voxel (i.e., 0.5-voxel cubic radius) to a maximum of 81 voxels (i.e., 2.5-voxel radius) across the task-related SPs with a size of 100 voxels for both the block-based and event-related task paradigms. In addition, several levels of maximum percentage BOLD intensity (i.e., 0.5, 1.0, 2.0, and 3.0%) were used for each degree of spatial overlap size. From the results, the estimated individual SPs of neuronal activation obtained from the proposed iterative DR approach with a sparse prior showed an enhanced true positive rate and reduced false positive rate compared to the ordinary DR approach. The estimated TCs of the task-related SPs from our proposed approach showed greater temporal correlation coefficients with a reference hemodynamic response function than those of the ordinary DR approach. Moreover, the efficacy of the proposed DR approach was also successfully demonstrated by the results of real fMRI data acquired from left-/right-hand clenching tasks in both block-based and event-related task paradigms., (Copyright © 2012 Elsevier Inc. All rights reserved.)

Intrinsic brain activity known as default-mode networks (DMNs) has been observed predominantly within the medial/superior frontal areas, anterior/posterior cingulate gyri, and precuneus using blood-oxygenation-level-dependent (BOLD) functional MRI (fMRI). Despite anecdotal evidence of distinct spatial patterns reflecting neuropsychiatric conditions in these DMNs, rigorous analysis of the characteristic traits of DMNs has been limited in previous studies. In this letter, the reproducibility and potential variability of the anterior and posterior DMNs were evaluated based on individual-level variations in effect sizes, activated areas, and causal interactions. Our results indicated that the DMNs were indeed reproducible between sessions/subjects. Region-specific traits were also observed: the posterior DMN seemed more robust to individual-level variations than the anterior DMN. The proposed analytical methods and reported findings may be useful in the development of a wide range of applications, including those involving clinical populations, which utilize the characteristic traits of DMNs., (Copyright © 2011 Elsevier Ireland Ltd. All rights reserved.)

The purpose of this study was to evaluate the diagnostic value of magnetic resonance imaging (MRI) and assess the correlation between the volume of the ischemic lesion and neurobehavioral status during the subacute stage of ischemic stroke. Ischemic stroke was induced in 6 healthy laboratory beagles through permanent occlusion of the middle cerebral artery (MCAO). T2-weighted and fluid-attenuated inversion recovery (FLAIR) imaging, diffusion-weighted imaging (DWI), measurement of the apparent diffusion coefficient (ADC) ratio, and neurobehavioral evaluation were performed 3 times serially by using a 1.5-T MR system: before and 3 and 10 d after MCAO. Ischemic lesions demonstrated T2 hyperintensity, FLAIR hyperintensity, and DWI hyperintensity. The ADC ratio was decreased initially but then was increased at 10 d after MCAO. Ischemic lesion volumes on T2-weighted and FLAIR imaging were not significantly different from those on DWI. The lesion volume and neurobehavioral score showed strong correlation. Our results suggest that conventional MRI may be a reliable diagnostic tool during the subacute stage of canine ischemic stroke.

Automated interpretation and classification of functional MRI (fMRI) data is an emerging research field that enables the characterization of underlying cognitive processes with minimal human intervention. In this work, we present a method for the automated classification of human thoughts reflected on a trial-based paradigm using fMRI with a significantly shortened data acquisition time (less than one minute). Based on our preliminary experience with various cognitive imagery tasks, six characteristic thoughts were chosen as target tasks for the present work: right-hand motor imagery, left-hand motor imagery, right foot motor imagery, mental calculation, internal speech/word generation, and visual imagery. These six tasks were performed by five healthy volunteers and functional images were obtained using a T(*)(2)-weighted echo planar imaging (EPI) sequence. Feature vectors from activation maps, necessary for the classification of neural activity, were automatically extracted from the regions that were consistently and exclusively activated for a given task during the training process. Extracted feature vectors were classified using the support vector machine (SVM) algorithm. Parameter optimization, using a k-fold cross validation scheme, allowed the successful recognition of the six different categories of administered thought tasks with an accuracy of 74.5% (mean)+/-14.3% (standard deviation) across all five subjects. Our proposed study for the automated classification of fMRI data may be utilized in further investigations to monitor/identify human thought processes and their potential link to hardware/computer control.

Real-time functional MRI (rtfMRI) has been used as a basis for brain-computer interface (BCI) due to its ability to characterize region-specific brain activity in real-time. As an extension of BCI, we present an rtfMRI-based brain-machine interface (BMI) whereby 2-dimensional movement of a robotic arm was controlled by the regulation (and concurrent detection) of regional cortical activations in the primary motor areas. To do so, the subjects were engaged in the right- and/or left-hand motor imagery tasks. The blood oxygenation level dependent (BOLD) signal originating from the corresponding hand motor areas was then translated into horizontal or vertical robotic arm movement. The movement was broadcasted visually back to the subject as a feedback. We demonstrated that real-time control of the robotic arm only through the subjects' thought processes was possible using the rtfMRI-based BMI trials.

The simultaneous acquisition of electroencephalogram (EEG) and functional MRI (fMRI) signals is potentially advantageous because of the superior resolution that is achieved in both the temporal and spatial domains, respectively. However, ballistocardiographic artifacts along with ocular artifacts are a major obstacle for the detection of the EEG signatures of interest. Since the sources corresponding to these artifacts are independent from those producing the EEG signatures, we applied the Infomax-based independent component analysis (ICA) technique to separate the EEG signatures from the artifacts. The isolated EEG signatures were further utilized to model the canonical hemodynamic response functions (HRFs). Subsequently, the brain areas from which these EEG signatures originated were identified as locales of activation patterns from the analysis of fMRI data. Upon the identification and subsequent evaluation of brain areas generating interictal epileptic discharge (IED) spikes from an epileptic subject, the presented method was successfully applied to detect the theta and alpha rhythms that are sleep onset-related EEG signatures along with the subsequent neural circuitries from a sleep-deprived volunteer. These results suggest that the ICA technique may be useful for the preprocessing of simultaneous EEG-fMRI acquisitions, especially when a reference paradigm is unavailable.

Independent component analysis (ICA) of fMRI data generates session/individual specific brain activation maps without a priori assumptions regarding the timing or pattern of the blood-oxygenation-level-dependent (BOLD) signal responses. However, because of a random permutation among output components, ICA does not offer a straightforward solution for the inference of group-level activation. In this study, we present an independent vector analysis (IVA) method to address the permutation problem during fMRI group data analysis. In comparison to ICA, IVA offers an analysis of additional dependent components, which were assigned for use in the automated grouping of dependent activation patterns across subjects. Upon testing using simulated trial-based fMRI data, our proposed method was applied to real fMRI data employing both a single-trial task-paradigm (right hand motor clenching and internal speech generation tasks) and a three-trial task-paradigm (right hand motor imagery task). A generalized linear model (GLM) and the group ICA of the fMRI toolbox (GIFT) were also applied to the same data set for comparison to IVA. Compared to GLM, IVA successfully captured activation patterns even when the functional areas showed variable hemodynamic responses that deviated from a hypothesized response. We also showed that IVA effectively inferred group-activation patterns of unknown origins without the requirement for a pre-processing stage (such as data concatenation in ICA-based GIFT). IVA can be used as a potential alternative or an adjunct to current ICA-based fMRI group processing methods.

We report an automated method to simultaneously monitor blood-oxygenation-level-dependent (BOLD) MR signals from multiple cortical areas in real-time. Individual brain anatomy was normalized and registered to a pre-segmented atlas in standardized anatomical space. Subsequently, using real-time fMRI (rtfMRI) data acquisition, localized BOLD signals were measured and displayed from user-selected areas labeled with anatomical and Brodmann's Area (BA) nomenclature. The method was tested on healthy volunteers during the performance of hand motor and internal speech generation tasks employing a trial-based design. Our data normalization and registration algorithm, along with image reconstruction, movement correction and a data display routine were executed with enough processing and communication bandwidth necessary for real-time operation. Task-specific BOLD signals were observed from the hand motor and language areas. One of the study participants was allowed to freely engage in hand clenching tasks, and associated brain activities were detected from the motor-related neural substrates without prior knowledge of the task onset time. The proposed method may be applied to various applications such as neurofeedback, brain-computer-interface, and functional mapping for surgical planning where real-time monitoring of region-specific brain activity is needed.

Background: Recent biomechanical studies have reported that stress on the posterior cruciate ligament (PCL) graft increases as the posterior tibial slope (PTS) decreases (flattened) in knees with single-bundle (SB) and double-bundle PCL reconstruction. Clinical studies of SB PCL reconstruction have shown that a flattened PTS is associated with a lesser reduction in posterior tibial translation. There is no long-term study on the clinical outcomes and graft survival rates of SB PCL reconstruction based on the medial and lateral PTSs measured on magnetic resonance imaging. Hypothesis: Flattened medial and lateral PTSs are associated with poor clinical outcomes and graft survival rates at a minimum 10-year follow-up after SB PCL reconstruction. Study Design: Cohort study; Level of evidence, 3. Methods: In this cohort study, we retrospectively reviewed 46 patients (mean age, 28.8 ± 9.9 years) who underwent primary SB PCL reconstruction between 2000 and 2009. They were followed up for a minimum of 10 years. The medial and lateral PTSs were measured on preoperative magnetic resonance imaging. As a previous study reported that a steeper medial or lateral PTS showed a higher risk of anterior tibial translation at thresholds of 5.6° and 3.8°, respectively, the patients were divided into 2 groups based on the cutoff values of both the medial (≤5.6° vs >5.6°) and lateral (≤3.8° vs >3.8°) PTSs. Clinical scores (International Knee Documentation Committee subjective score, Lysholm score, and Tegner activity score), radiological outcomes (side-to-side difference [SSD] on stress radiography and osteoarthritis progression), and graft survival rates were compared between the groups at the last follow-up. Results: All clinical scores and the progression of osteoarthritis demonstrated no significant difference between the 2 subgroups of both the medial and lateral PTS groups. The mean SSD on stress radiography after SB PCL reconstruction was significantly greater in patients with a medial PTS ≤5.6° than in patients with a medial PTS >5.6° (8.4 ± 3.9 vs 5.1 ± 2.9 mm, respectively; P =.030), while the lateral PTS subgroups after SB PCL reconstruction demonstrated no significant difference. The minimum 10-year graft survival rate was significantly lower in patients with a medial PTS ≤5.6° (68.4% vs 92.6%, respectively; P =.029) and a lateral PTS ≤3.8° (50.0% vs 91.7%, respectively; P =.001). Conclusion: A flattened medial PTS (≤5.6°) was associated with an increased SSD on stress radiography, and both flattened medial (≤5.6°) and lateral (≤3.8°) PTSs resulted in lower graft survival rates at a minimum 10-year follow-up after primary SB PCL reconstruction. [ABSTRACT FROM AUTHOR]

Background: Anatomic anterior cruciate ligament reconstruction (ACLR) is preferred over nonanatomic ACLR. However, there is no consensus on which point the tunnels should be positioned among the broad anatomic footprints. Purpose/Hypothesis: To identify the ideal combination of tibial and femoral tunnel positions according to the femoral and tibial footprints of the anteromedial (AM) and posterolateral (PL) anterior cruciate ligament bundles. It was hypothesized that patients with anteromedially positioned tunnels would have better clinical scores, knee joint stability, and graft signal intensity on follow-up magnetic resonance imaging (MRI) than those with posterolaterally positioned tunnels. Study Design: Cohort study; Level of evidence, 3. Methods: A total of 119 patients who underwent isolated single-bundle ACLR with a hamstring autograft from July 2013 to September 2018 were retrospectively investigated. Included were patients with clinical scores and knee joint stability test results at 2-year follow-up and postoperative 3-dimensional computed tomography and 1-year postoperative MRI findings. The cohort was divided into 4 groups, named according to the bundle positions in the tibial and femoral tunnels: AM-AM (n = 33), AM-PL (n = 26), PL-AM (n = 29), and PL-PL (n = 31). Results: There were no statistically significant differences among the 4 groups in preoperative demographic data or postoperative clinical scores (Lysholm, Tegner, and International Knee Documentation Committee subjective scores); knee joint stability (anterior drawer, Lachman, and pivot-shift tests and Telos stress radiographic measurement of the side-to-side difference in anterior tibial translation); graft signal intensity on follow-up MRI; or graft failure. Conclusion: No significant differences in clinical scores, knee joint stability, or graft signal intensity on follow-up MRI were identified between the patients with anteromedially and posterolaterally positioned tunnels. [ABSTRACT FROM AUTHOR]

The cryogenic pump is a crucial component of the magnetic resonance imaging (MRI) system for delivering liquid helium to a magnet for superconductivity, thereby generating a mechanical vibration. Thus, the cryogenic pump for liquid helium (helium pump) contaminates the electroencephalography (EEG) simultaneously acquired with functional MRI. The recursive approach of EEG‐segment‐based principal component analysis (rsPCA) has recently demonstrated its efficacy in removing this helium pump artefact. In the rsPCA, the recursion depth and EEG‐segment size are crucial hyperparameters. rsPCA's performance and computational time across recursion depth (1, 2, 3) and segment sizes (165, 220, 265) are systematically evaluated. It is found that the recursion depth of 2 yielded significant reductions in the computational time compared to the depth of 3 across all segment sizes while maintaining the denoising performance. The binary classification performance (left‐hand versus right‐hand clenching) was also enhanced in this scenario (especially the use of EEG‐segment size of 165 and 220) by using EEG gamma‐band activity (30–50 Hz), which is predominantly contaminated by the helium‐pump artefact. [ABSTRACT FROM AUTHOR]

Background: Brain activation differs according to lesion location in functional magnetic resonance imaging (fMRI) studies, but lesion location-dependent electroencephalographic (EEG) alterations are unclear. Because of the increasing use of EEG-based brain-computer-interface rehabilitation, we examined lesion location-dependent EEG patterns in patients with stroke while they performed motor tasks.Methods: Twelve patients with chronic stroke were divided into three subgroups according to their lesion locations: supratentorial lesions that included M1 (SM1+), supratentorial lesions that excluded M1 (SM1-), and infratentorial (INF) lesions. Participants performed three motor tasks [active, passive, and motor imagery (MI)] with supination and grasping movements. The hemispheric asymmetric indexes, which were calculated with laterality coefficients (LCs), the temporal changes in the event-related desynchronization (ERD) patterns in the bilateral motor cortex, and the topographical distributions in the 28-channel EEG patterns around the supplementary motor area and bilateral motor cortex of the three participant subgroups were compared with those of the 12 age-matched healthy controls.Results: The SM1+ group exhibited negative LC values in the active and MI motor tasks, while the other patient subgroups exhibited positive LC values. Negative LC values indicate that the ERD/ERS intensity of the ipsilateral hemisphere is higher than the contralateral hemisphere, whereas positive LC values indicate that the ERD/ERS intensity of the contralateral hemisphere is higher than the ipsilateral hemisphere. The LC values of SM1+ and healthy controls differed significantly (rank-sum test, p < 0.05) in both the supination and grasping movements in the active task. The three patient subgroups differed distinctly from each other in the topography analysis.Conclusions: The hemispheric asymmetry and topographic characteristics of the beta band power patterns in the patients with stroke differed according to the location of the lesion, which suggested that EEG analyses of neurorehabilitation should be implemented according to lesion location. [ABSTRACT FROM AUTHOR]

Spontaneous fluctuations of the functional magnetic resonance imaging (fMRI) signal during the resting state have been characterized by the dominant spectral components in the low-frequency (<0.1 Hz) range. Although previous studies have reported the spatial patterns of activity of the resting-state network (RSN), the frequency-dependent characteristics (i.e., spectral coherence) of the resting state activity have not been fully investigated. This article describes the novel application of a non-negative matrix factorization (NMF) algorithm to the decomposition of the magnitude spectra of fMRI time-series into distinct spectral components. From the fMRI data of healthy volunteers during the resting state, the frequency-specific components were decomposed into five basis functions. Group analysis revealed five different spatial patterns associated with these basis functions, in which each of the spatial patterns may correspond to a distinct spectrally coherent RSN. The RSN with the lowest center frequency showed a similar spatial pattern to the 'default-mode' network, which involves the medial superior and middle frontal cortex along with the posterior cingulate cortex. On the other hand, RSNs with higher frequencies were observed mainly in several posterior regions of the brain including the precuneus and lingual gyrus. Subsequent Granger causality analysis demonstrated that these posterior regions may function as the 'hubs' of the RSNs, whereas the anterior regions, including the medial superior and middle frontal cortex, may be characterized as 'peripheries' of the network. Our proposed analysis scheme provides supplemental information on both spectral and temporal characteristics of the RSNs, which might be used in a range of applications, including those involving clinical populations. © 2011 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 21, 2117-222, 2011 [ABSTRACT FROM AUTHOR]

The investigation of the reproducibility in functional MRI (fMRI) is an important step in the quantification and analysis of paradigm-related brain activation. This article reports on reproducibility of cortical activation characterized by repeated fMRI runs (10 times) during the performance of a motor imagery and a passive auditory stimulation as a control task. Two parameters, the size of activation and BOLD signal contrast, were measured from regions-of-interest for 10 subjects across different threshold conditions. The variability of these parameters was normalized with respect to the mean obtained from 10 runs, and represented as the intrasession variability. It was found that the variability was significantly lower in the measurement of BOLD signal contrast as compared to the measurement of the size of activation. The variability of the activation volume measurement was greater in the motor imagery task than in the auditory tasks across all thresholds. This task-dependent difference was not apparent from the measurement of the BOLD signal contrast. The presence of threshold dependence in the variability measurement was also examined, but no such dependency was found. The results suggest that a measurement of BOLD signal itself is a more reliable indicator of paradigm-related brain activation during repeated fMRI scans. [ABSTRACT FROM AUTHOR]

Abstract An artificial neural network with multiple hidden layers (known as a deep neural network, or DNN) was employed as a predictive model (DNN p) for the first time to predict emotional responses using whole-brain functional magnetic resonance imaging (fMRI) data from individual subjects. During fMRI data acquisition, 10 healthy participants listened to 80 International Affective Digital Sound stimuli and rated their own emotions generated by each sound stimulus in terms of the arousal, dominance, and valence dimensions. The whole-brain spatial patterns from a general linear model (i.e., beta-valued maps) for each sound stimulus and the emotional response ratings were used as the input and output for the DNN P , respectively. Based on a nested five-fold cross-validation scheme, the paired input and output data were divided into training (three-fold), validation (one-fold), and test (one-fold) data. The DNN P was trained and optimized using the training and validation data and was tested using the test data. The Pearson's correlation coefficients between the rated and predicted emotional responses from our DNN P model with weight sparsity optimization (mean ± standard error 0.52 ± 0.02 for arousal, 0.51 ± 0.03 for dominance, and 0.51 ± 0.03 for valence, with an input denoising level of 0.3 and a mini-batch size of 1) were significantly greater than those of DNN models with conventional regularization schemes including elastic net regularization (0.15 ± 0.05, 0.15 ± 0.06, and 0.21 ± 0.04 for arousal, dominance, and valence, respectively), those of shallow models including logistic regression (0.11 ± 0.04, 0.10 ± 0.05, and 0.17 ± 0.04 for arousal, dominance, and valence, respectively; average of logistic regression and sparse logistic regression), and those of support vector machine-based predictive models (SVM p s; 0.12 ± 0.06, 0.06 ± 0.06, and 0.10 ± 0.06 for arousal, dominance, and valence, respectively; average of linear and non-linear SVM p s). This difference was confirmed to be significant with a Bonferroni-corrected p -value of less than 0.001 from a one-way analysis of variance (ANOVA) and subsequent paired t -test. The weights of the trained DNN P s were interpreted and input patterns that maximized or minimized the output of the DNN P s (i.e., the emotional responses) were estimated. Based on a binary classification of each emotion category (e.g., high arousal vs. low arousal), the error rates for the DNN P (31.2% ± 1.3% for arousal, 29.0% ± 1.7% for dominance, and 28.6% ± 3.0% for valence) were significantly lower than those for the linear SVM P (44.7% ± 2.0%, 50.7% ± 1.7%, and 47.4% ± 1.9% for arousal, dominance, and valence, respectively) and the non-linear SVM P (48.8% ± 2.3%, 52.2% ± 1.9%, and 46.4% ± 1.3% for arousal, dominance, and valence, respectively), as confirmed by the Bonferroni-corrected p < 0.001 from the one-way ANOVA. Our study demonstrates that the DNN p model is able to reveal neuronal circuitry associated with human emotional processing – including structures in the limbic and paralimbic areas, which include the amygdala, prefrontal areas, anterior cingulate cortex, insula, and caudate. Our DNN p model was also able to use activation patterns in these structures to predict and classify emotional responses to stimuli. Highlights • Deep neural network (DNN) was trained to predict the human emotion measured from fMRI. • The prediction performance of the DNN was superior to that of the support vector machine. • Weight representation and input pattern estimation were introduced to interpret the trained DNN. • Brain regions related to emotional processing were identified from the DNN interpretation. • Representations of neuronal activations were readily separable at the higher hidden layer of the DNN. [ABSTRACT FROM AUTHOR]

Abstract: Background: Women have approximately twice the risk of major depressive disorder (MDD) than men, yet this difference remains largely unexplained. Previous MDD research suggests high rates of endocrine dysfunction, which may be related to deficits in brain activity in stress response circuitry [hypothalamus, amygdala, hippocampus, anterior cingulate cortex (ACC), orbitofrontal cortex (OFC)]. This functional magnetic resonance imaging (fMRI) study investigated the relationship between hypothalamic–pituitary–gonadal (HPG)-axis hormones and stress response circuitry dysfunction in MDD in women. Methods: During the late follicular/midcycle phase of the menstrual cycle, female participants (10 with extensive histories of MDD, in remission, 10 healthy controls) were scanned while viewing negative and neutral arousal pictures. Group differences in blood oxygen-level dependent (BOLD) signal changes were analyzed using SPM2. Baseline gonadal hormones included estradiol, progesterone, and testosterone. Results: fMRI results showed greater BOLD signal intensity changes in controls versus MDD in hypothalamus, amygdala, hippocampus, OFC, ACC, and subgenual ACC, findings unrelated to medication status. MDD women had a lower serum estradiol and higher serum progesterone compared to controls. Hypoactivations in hypothalamus, subgenual ACC, amygdala and OFC in MDD were associated with low estradiol and high progesterone. Limitations: Generalizability of our findings is limited by small sample size and restriction to females, although this did not affect the internal validity of the results. Conclusions: Hypoactivation of the stress response circuitry in MDD women is associated with dysregulation of the HPG-axis. Associations between brain activity deficits and hormonal disruption in MDD may ultimately contribute to understanding sex differences in MDD. [Copyright &y& Elsevier]