Generating Synthetic Task-based Brain Fingerprints for Population Neuroscience Using Deep Learning

Epilepsy affects neural processing and often causes intra- or interhemispheric language reorganization, rendering localization solely based on anatomical landmarks (e.g., Broca's area) unreliable. Preoperative brain mapping is necessary to weigh the risk of resection with the risk of postoperative deficit. However, the use of conventional mapping methods (e.g., somatosensory stimulation, task-based functional MRI [fMRI]) in pediatric patients is technically difficult due to low compliance and their unique neurophysiology. Resting-state fMRI (rs-fMRI), a "task-free" technique based on the neural activity of the brain at rest, has the potential to overcome these limitations. The authors hypothesized that language networks can be identified from rs-fMRI by applying functional connectivity analyses. Cases in which both task-based fMRI and rs-fMRI were acquired as part of the preoperative clinical protocol for epilepsy surgery were reviewed. Task-based fMRI consisted of 2 language tasks and 1 motor task. Resting-state fMRI data were acquired while the patients watched an animated movie and were analyzed using independent component analysis (i.e., data-driven method). The authors extracted language networks from rs-fMRI data by performing a similarity analysis with functionally defined language network templates via a template-matching procedure. The Dice coefficient was used to quantify the overlap. Thirteen children underwent conventional task-based fMRI (e.g., verb generation, object naming), rs-fMRI, and structural imaging at 1.5T. The language components with the highest overlap with the language templates were identified for each patient. Language lateralization results from task-based fMRI and rs-fMRI mapping were comparable, with good concordance in most cases. Resting-state fMRI-derived language maps indicated that language was on the left in 4 patients (31%), on the right in 5 patients (38%), and bilateral in 4 patients (31%). In some cases, rs-fMRI indicated a more extensive language representation. Resting-state fMRI-derived language network data were identified at the patient level using a template-matching method. More than half of the patients in this study presented with atypical language lateralization, emphasizing the need for mapping. Overall, these data suggest that this technique may be used to preoperatively identify language networks in pediatric patients. It may also optimize presurgical planning of electrode placement and thereby guide the surgeon's approach to the epileptogenic zone.

This study investigates the dynamic functional con-nectivity (DFC) in resting-state fMRI (rs-fMRI) data of chess players using a Vector Auto-Regression (VAR) model. The VAR model was constructed using the Group Lasso and Sliding Window technique. The study included 116 brain regions, and their correlation was examined in the context of their dynamic connection. Statistical feature selection techniques were used to determine which dynamic connections of brain areas were crucial in discriminating chess masters from novice players. After identifying key DFCs related to these brain regions, a classification model was built to classify chess experts and normal control individuals.Our classification model achieved an accuracy of 96.33%under a 10-fold cross-validation framework. This performance represents a substantial improvement over previous studies uti-lizing only rs-fMRI data, which reported a maximum accuracy of 85.45%, indicating a 10.88% enhancement in accuracy. Moreover, our model outperformed methods that combined rs-fMRI with T1-weighted MRI data, which achieved an accuracy of 88%, yielding an additional 8.33% improvement. These results demon-strate that our approach, relying solely on rs-fMRI data, offers a notable advancement in the classification of chess expertise.

Of Bandwagons and Bathwater: The Value of Resting State Functional Magnetic Resonance Imaging for Child Psychiatric Research

Several previous attempts have been made to divide the human amygdala into smaller subregions based on the unique functional properties of the subregions. Although these attempts have provided valuable insight into the functional heterogeneity in this structure, the possibility that spatial patterns of functional characteristics can quickly change over time has rarely been considered in previous studies. In the present study, we explicitly account for the dynamic nature of amygdala activity. Our goal was not only to develop another parcellation method but also to augment existing methods with novel information about amygdala subdivisions. We performed state-specific amygdala parcellation using resting-state fMRI (rsfMRI) data and recurrence quantification analysis (RQA). RsfMRI data from 102 subjects were acquired with a 3T Trio Siemens scanner. We analyzed values of several RQA measures across all voxels in the amygdala and found two amygdala subdivisions, the ventrolateral (VL) and dorsomedial (DM) subdivisions, that differ with respect to one of the RQA measures, Shannon's entropy of diagonal lines. Compared to the DM subdivision, the VL subdivision can be characterized by a higher value of entropy. The results suggest that VL activity is determined and influenced by more brain structures than is DM activity. To assess the biological validity of the obtained subdivisions, we compared them with histological atlases and currently available parcellations based on structural connectivity patterns (Anatomy Probability Maps) and cytoarchitectonic features (SPM Anatomy toolbox). Moreover, we examined their cortical and subcortical functional connectivity. The obtained results are similar to those previously reported on parcellation performed on the basis of structural connectivity patterns. Functional connectivity analysis revealed that the VL subdivision has strong connections to several cortical areas, whereas the DM subdivision is mainly connected to subcortical regions. This finding suggests that the VL subdivision corresponds to the basolateral subdivision of the amygdala (BLA), while the DM subdivision has some characteristics typical of the centromedial amygdala (CMA). The similarity in functional connectivity patterns between the VL subdivision and BLA, as well as between the DM subdivision and CMA, confirm the utility of our parcellation method. Overall, the study shows that parcellation based on BOLD signal dynamics is a powerful tool for identifying distinct functional systems within the amygdala. This tool might be useful for future research on functional brain organization.

Neurodevelopmental conditions such as Autism Spectrum Disorder (ASD) and Attention Deficit Hyperactivity Disorder (ADHD) alter functional connectivity (FC) among the brain regions. ADHD and ASD affect <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{5 - 6 \%}$</tex> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{1 - 2 \%}$</tex> of the population, respectively and hence need immediate attention to develop automated differential diagnosis. This work proposes an automated diagnostic system for ADHD and ASD by investigating FC among brain regions. We used publicly available restingstate functional Magnetic Resonance Imaging (rs-fMRI) data for ADHD, ASD and age-matched healthy (HC) subjects. The rsfMRI data helps to analyse the temporal activity in the brain, and computing FC using time-series data helps to capture the temporal correlation among regions. We used five diverse FC measures and performed three classification tasks: ASD vs. ADHD, ASD vs. HC and ADHD vs. HC. Further, we finetuned the pre-trained VGG19 model to develop a decision model. The best accuracy of 99.09% for ASD vs. ADHD, 78.18% for ADHD vs. HC and 65.46% for ASD vs. HC demonstrates the efficacy of the proposed method. The disrupted regions are also identified using SHapley Additive exPlanations (SHAP), an explainable artificial intelligence method, that are in accordance with the literature. The work demonstrates the capability of deep learning in analysing neuroimaging data for clinical and research applications.

BackgroundExcessive daytime sleepiness (EDS) is a frequent nonmotor symptoms of Parkinson’s disease (PD), which seriously affects the quality of life of PD patients and exacerbates other nonmotor symptoms. Previous studies have used static analyses of these resting-state functional magnetic resonance imaging (rs-fMRI) data were measured under the assumption that the intrinsic fluctuations during MRI scans are stationary. However, dynamic functional network connectivity (dFNC) analysis captures time-varying connectivity over short time scales and may reveal complex functional tissues in the brain.PurposeTo identify dynamic functional connectivity characteristics in PD-EDS patients in order to explain the underlying neuropathological mechanisms.MethodsBased on rs-fMRI data from 16 PD patients with EDS and 41 PD patients without EDS, we applied the sliding window approach, k-means clustering and independent component analysis to estimate the inherent dynamic connectivity states associated with EDS in PD patients and investigated the differences between groups. Furthermore, to assess the correlations between the altered temporal properties and the Epworth sleepiness scale (ESS) scores.ResultsWe found four distinct functional connectivity states in PD patients. The patients in the PD-EDS group showed increased fractional time and mean dwell time in state IV, which was characterized by strong connectivity in the sensorimotor (SMN) and visual (VIS) networks, and reduced fractional time in state I, which was characterized by strong positive connectivity intranetwork of the default mode network (DMN) and VIS, while negative connectivity internetwork between the DMN and VIS. Moreover, the ESS scores were positively correlated with fraction time in state IV.ConclusionOur results indicated that the strong connectivity within and between the SMN and VIS was characteristic of EDS in PD patients, which may be a potential marker of pathophysiological features related to EDS in PD patients.

NBS-Predict: A prediction-based extension of the network-based statistic

Optimization of rs-fMRI Pre-processing for Enhanced Signal-Noise Separation, Test-Retest Reliability, and Group Discrimination

Resting-state functional MRI (rs-fMRI) is a well-established method for studying intrinsic connectivity and mapping the topography of functional networks in the human brain. In the clinical setting, rs-fMRI has been used to define functional topography, typically language and motor systems, in the context of preoperative planning for neurosurgery. Intraoperative mapping of critical speech and motor areas with electrocortical stimulation (ECS) remains standard practice, but preoperative noninvasive mapping has the potential to reduce operative time and provide functional localization when awake mapping is not feasible. Task-based fMRI has historically been used for this purpose, but it can be limited by the young age of the patient, cognitive impairment, poor cooperation, and need for sedation. Resting-state fMRI allows reliable analysis of all functional networks with a single study and is inherently independent of factors affecting task performance. In this review, the authors provide a summary of the theory and methods for resting-state network mapping. They provide case examples illustrating clinical implementation and discuss limitations of rs-fMRI and review available data regarding performance in comparison to ECS. Finally, they discuss novel opportunities for future clinical applications and prospects for rs-fMRI beyond mapping of regions to avoid during surgery but, instead, as a tool to guide novel network-based therapies.

In recent years, several functional imaging studies have shown that the hypothalamus is closely associated with migraine and have suggested that the hypothalamus may be a potential site of migraine generation. Studying the characteristics of the functional network of the hypothalamus in persons with migraines may help to understand the neural mechanisms of migraine. We thus used resting-state functional magnetic resonance imaging (rsfMRI) and Granger causal analysis to investigate the effective connectivity (EC) of the hypothalamus in persons with migraines. The study included 17 healthy volunteers and 39 persons with migraines. The EC calculation was based on rsfMRI data from a 3-T magnetic resonance imaging scanner. The brain networks of the hypothalamus were compared using a general linear model to determine if there were any differences between the two groups. We used Pearson correlation analysis to examine the correlation between EC values in abnormal brain regions and clinical variables. Compared with healthy controls, those with migraines showed decreased EC from the hypothalamus to the left fusiform and increased EC from the hypothalamus to the medial frontal gyrus/orbital part, right lingual gyrus, left superior frontal gyrus, and right middle frontal gyrus (P<0.05). Meanwhile, persons with migraines also showed decreased EC from the left middle frontal gyrus and right medial frontal gyrus/orbital part to the hypothalamus (P<0.05). EC from the hypothalamus to the left superior frontal gyrus correlated significantly and positively with the visual analogue scale in those with migraines (r=-0.3820; P=0.0164). Disturbances in the EC between the hypothalamus and the prefrontal gyrus and visual cortex may play a key role in the neuropathological features of persons with migraines. The current study adds to our understanding of the complexity of migraine mechanisms.

BackgroundThis study aimed to investigate the correlations between abnormal spontaneous neural activity measured with fractional amplitude of low-frequency fluctuations (fALFF) and clinical variables, eye movements, and event-related potential indicators in patients with major depressive disorder (MDD).MethodsWe recruited 42 patients with MDD and 42 healthy controls (HCs) and collected their clinical variables, eye movement, event-related potential, and resting-state functional magnetic resonance imaging (rs-fMRI) data. The fALFF, support vector machine (SVM), and correlation analysis were used to analyze the data.ResultsThe results of the study showed that the fALFF values of the sensorimotor network, including the right middle temporal gyrus, right cerebellar Crus2, left occipital gyrus, and left middle temporal gyrus, were significantly higher compared to HCs. Correlation analysis showed that the abnormal fALFF value of the right cerebellar Crus2 was inversely correlated with the active coping scores of the Simplified Coping Style Questionnaire in the patients (r = −0.307, p = 0.048). No correlation was observed between abnormal fALFF values and other clinical symptoms, neuropsychological tests, eye movements, and event-related potential-related indicators in patients with MDD. fALFF values in the left middle temporal gyrus could be used to distinguish patients with MDD from HCs with an accuracy of 78.57%.ConclusionsPatients with MDD exhibited enhanced spontaneous neural activity in the sensorimotor network. No associations were found between abnormal spontaneous neural activity and clinical variables, eye movements, and event-related potential related indicators in MDD.

Major Depressive Disorder (MDD) significantly impacts global health, impairing individual functioning and increasing socioeconomic burden. Developing innovative, interpretable approaches for its identification is essential for improving diagnosis and guiding treatment. This study introduces a novel framework designed to classify MDD using resting-state functional MRI (rs-fMRI) data. Our framework follows three stages: First, Node2Vec extracts rich, low-dimensional brain region embeddings from functional connectivity (FC) networks, capturing their complex topological information. Second, these informative embeddings feed a Graph Attention Network (GAT) which, via multi-head attention, identifies and weighs discriminative inter-regional functional connections, refining them into a potent graph representation. Third, these GAT-derived representations are classified by an ensemble classifier (Random Forest, SVM, MLP) for robust MDD identification. The model achieved classification accuracies of 78.73% and 92.94% on the REST-meta-MDD and SRPBS-MDD datasets, respectively. Moreover, the attention mechanism revealed that resting-state functional connectivity of regions within the Default Mode Network (DMN) and Frontoparietal Network (FPN) were among the most discriminative features for distinguishing MDD from healthy controls. The attention mechanism enhanced interpretability by highlighting significant brain regions linked to MDD. Compared to traditional methods, this GNN-based approach effectively captures complex brain connectivity patterns and offers improved interpretability, ultimately aiding healthcare professionals in diagnosing MDD more accurately.

Recent studies have highlighted the importance of analyzing spectral power in resting-state functional magnetic resonance imaging (rs-fMRI) data. Significant modulation of power has been ascribed to the performance of cognitive tasks and has been ascribed clinical significance. However, the role of confounding factors such as head motion on spectral power is not fully understood. Specifically, the spatial distribution of frequency-dependent associations between rs-fMRI power and motion is unknown. We utilized a large rs-fMRI dataset (n=1000) to quantify the influence of head motion on spectral power in different frequency bands. We (1) performed regression analyses across the entire sample and (2) computed difference maps between high- and low-motion groups, more consistent with common experimental designs, and both analyses gave similar results. Greater head motion led to reduced spectral power at lower frequencies (0.007-0.05 Hz), but increased power at higher frequencies (0.12-0.167 Hz). Importantly, our whole-brain voxel-wise analysis showed that brain areas in distributed association networks (e.g., default mode and frontoparietal control networks) were most susceptible to head motion. These results were consistent with or without global signal regression (GSR). Additionally, without GSR, we noted a positive correlation with low-frequency power in the pre- and postcentral gyrus (S1/M1), mid-cingulate cortex, and insula and a negative correlation with mid-frequency (0.05-0.12 Hz) power in S1/M1, visual, and lateral temporal cortices. Hence, head motion significantly affects rs-fMRI power and great care must be taken when assigning a diagnostic marker for clinical populations known to present with greater head motion.

Sensory, somatomotor and internal mentation networks emerge dynamically in the resting brain with internal mentation predominating in older age

Traumatic brain injury (TBI) is associated with widespread disruptions in functional connectivity (FC), yet how these alterations vary by injury severity remains unclear. Traditional classification systems fail to capture network-level dysfunction, limiting prognostic accuracy and targeted rehabilitation strategies. The aim of this study was to systematically evaluate fMRI-detected FC alterations after mild, moderate-severe, and severe TBI using coordinate-based meta-analysis and network-level mapping. A systematic search of MEDLINE/PubMed, Embase, and Web of Science was conducted to identify studies examining FC changes in TBI using fMRI. This review was not funded or prospectively registered. Studies were stratified by TBI severity and time since injury. Significant peak Montreal Neurological Institute coordinates were extracted, matched to the Yeo-17 brain network atlas, and analyzed using Seed-based d Mapping with Permutation of Subject Images (SDM-PSI). Study quality and evidence level were assessed using an adapted NIH Quality Assessment Tool and the Oxford Centre for Evidence-Based Medicine criteria. Eligible studies included adult participants with TBI assessed using resting-state or task-based fMRI; studies lacking severity classification or involving pediatric populations were excluded. Seventy-six studies were included, totaling 5,064 participants (2,993 patients with TBI, 1,914 controls; mean age 35.5 vs 35.0 years; 37.6% female overall). Mild TBI (mTBI) was the most common severity (n = 59 studies; 77.6%). Twenty-two studies contributed data for meta-analysis: 15 with resting-state fMRI and 7 with task-based fMRI. Aggregated peak coordinates most frequently mapped to subcomponents of the default mode (22.9%), ventral attention (18.8%), and somatomotor (10.1%) networks in mTBI and to the frontoparietal (36%), ventral attention (20%), and dorsal attention (14.7%) networks in moderate-severe/severe TBI. SDM-PSI identified uncorrected clusters in default mode and frontoparietal regions in mTBI and moderate-severe/severe TBI, respectively, but no clusters survived family-wise error rate correction (standardized mean difference Z score range -1.986 to 3.911, p < 0.05 uncorrected). Heterogeneity was low across analyses (I2 < 21%). FC changes after TBI potentially involve large-scale brain networks such as the default mode, attention, and executive control networks in a severity-dependent and phase-dependent manner. Although meta-analysis revealed consistent patterns, corrected statistical significance was not achieved, highlighting the need for larger, harmonized data sets and standardized analysis pipelines in future research.