NeuroImage

Voxelwise encoding models trained on functional MRI data can produce detailed maps of cortical organization. However, voxelwise encoding models must be trained on many hours of brain responses from each participant, limiting clinical applications. In this study, we introduce a cross-participant modeling framework for rapid cortical mapping. In this framework, voxelwise encoding models are trained on many hours of brain responses from previously scanned reference participants, and then transferred to a new participant by aligning brain responses using a small set of stimuli. We evaluated cross-participant encoding models on linguistic semantic mapping, non-linguistic semantic mapping, and auditory mapping. In each case we found that cross-participant encoding models had more accurate selectivity estimates and prediction performance than within-participant encoding models trained on the same amount of data from the new participant. We also found that cross-participant encoding models improved with the amount of data from each reference participant and the number of reference participants. These results demonstrate that cross-participant modeling can substantially reduce the amount of data required for accurate cortical mapping, which may facilitate new clinical applications of functional neuroimaging.

Threshold-free cluster enhancement (TFCE) is one of the most used statistical inference methods in neuroimaging, but its computational cost limits some of its applications. The current implementations recompute clusters at each threshold step, creating computation costs that poorly scale with precision increases. Furthermore, as larger samples and reduced noise increase maximum t-statistics, computational burden grows correspondingly. As the field moves towards finer parcellations, the number of FC edges grows quadratically with the number of ROIs, making TFCE computationally infeasible at the scales increasingly demanded by the field. We present Incremental Cluster TFCE (IC-TFCE), an algorithm that produces numerically equivalent results to standard TFCE while decoupling runtime from discretization precision. The IC-TFCE builds clusters incrementally from previous threshold steps rather than recomputing them, stores TFCE results on a region of interest (ROI) based structure instead of a functional connectivity (FC) edge structure for improved speed, and can be applied to voxel data through a novel graph transformation described and validated herein. This algorithm achieves a measured 3-93x speedup for FC TFCE depending on the precision parameter $dh$, making TFCE analyses with fine parcellations of 1000 or more ROIs computationally tractable for the first time. Finally, we validate correctness through mathematical proof and numerical comparison. The efficiency provided by IC-TFCE allowed a large-scale empirical power analysis across $dh$ values to guide practitioners in parameter selection for their analyses.

The brains functional organization has been extensively studied through pairwise connectivity analyses. While these approaches have provided important insights into brain network organization, they fall short in capturing the complexity of high-order functional interactions (HOI). Particularly relevant is the investigation of redundancy and synergy patterns -not addressable with pairwise interactions-, revealing fundamental mechanisms of brain integration and information processing across various cognitive functions and clinical conditions. Conventional neuroimaging software packages are primarily designed for classical (general linear model-like) analyses but lack native support for HOI metrics. To address this gap, this study introduces a novel framework that bridges high-order information theory with conventional neuroimaging analysis pipelines and is subsequently applied to resting-state functional MRI to demonstrate its practical utility. By representing HOI into voxel-level metrics, our approach allows standard neuroimaging analyses to probe complex multivariate dependencies. Moreover, voxel-level group-comparison analyses show age differences linked with reduced redundancy in default mode network interactions. These findings advance our understanding of the complex relationship between multivariate functional interactions, voxel-level neuroimaging, and behavior, highlighting novel analytic strategies to study high-order information processing underlying cognitive function and its alterations in pathological conditions.

Resting-state functional MRI (rs-fMRI) is a cornerstone of human brain research, yet its interpretation is complicated by its sensitivity to the slow hemodynamic response that obscures the organization of neural activity across faster time scales. Here we use simultaneous wide-field optical imaging (WOI) and rs-fMRI to directly examine the relationship between neural and hemodynamic functional connectomes across time scales. We show that much of the large-scale spatial structure is preserved across modalities, across time scales, and across frequencies. Although rs-fMRI robustly captures time-averaged neural activity, time-resolved rs-fMRI estimates of functional connectivity exhibit significantly greater variability, which partially reflects sensitivity limitations. Hemodynamic WOI signals maintain greater similarity to neural activity than rs-fMRI, although their fidelity is reduced at high frequencies. Together, our findings demonstrate that the time-averaged spatial structure of neural activity is faithfully represented in hemodynamics and rs-fMRI; provide insight into the reliability of time-resolved rs-fMRI across temporal scales; and establish a multimodal framework for validating features of dynamic brain activity. TeaserSpatial patterns of neural activity present across time scales are largely preserved in hemodynamics measured with optical imaging and rs-fMRI.

The human neocortex is organized into six laminae forming the structural basis for feedforward and feedback connections across the brain, yet their functional contributions have remained largely inaccessible for non-invasive imaging methods. Leveraging the ultrahigh field of a 10.5 Tesla scanner, we acquired anatomical and functional MRI data at 0.37mm ([~]0.05 {micro}L) and 0.35mm ([~]0.04 {micro}L) isotropic resolution, respectively, approaching the scale of individual cortical layers in humans. Using the Stria of Gennari as an in-vivo anatomical landmark, we extend our previous finding that feedforward visual activation in layer IV of the primary visual cortex during visual stimulation was resolved in laminar BOLD profiles. These laminar features were reproducible across sessions and were not clearly visible with more typical 0.8 mm resolutions at 7T, underscoring the benefits of further increases in magnetic field strength and resolution. This imaging domain, however, comes with increasing challenges of distortion, alignment, and cortical depth estimation, which must be addressed and mitigated to realize its benefits. In this paper we discuss the promises and challenges of this new regime of high resolutions. Our findings showcase the potential of ultrahigh field, ultrahigh resolution human fMRI to bridge the gap with invasive imaging of cortical layers.

Portable low-field MRI offers an affordable and mobile alternative to conventional high-field scanners, enabling imaging in point-of-care and resource-limited settings. However, its lower signal-to-noise ratio, reduced resolution, and acquisition artifacts raise concerns about the accuracy of standard image registration methods. Reliable registration is critical for a wide range of emerging applications, including frequent brain monitoring, assessment of neurodegenerative disease progression, and evaluation of treatment effects such as those of Alzheimers therapeutics. In this work, we systematically evaluated state-of-the-art registration approaches on simulated low-field scans (obtained by downsampling high-field images) and on real low-field brain MRI data. We compared three representative approaches: classical optimization (NiftyReg), learning-based registration (SynthMorph), and synthesis-based registration (SynthSR+NiftyReg). Using downsampled high-field scans, all methods performed well, achieving high Dice scores and smooth deformation fields, indicating that reduced resolution alone does not hinder registration. In contrast, real low-field data exhibited lower accuracy, primarily due to geometric distortion and other acquisition-specific artifacts. Among the tested approaches, the synthesis-based pipeline achieved the most robust performance across subjects and modalities. Overall, existing algorithms can accommodate resolution limitations, however, future methods could further enhance coregistration by explicitly addressing the distortions present in low-field MRI scans.

Magnetoencephalography (MEG) forward and inverse modeling is fundamental to neuroscientific discovery, yet the inversion of partial differential equations (PDEs) remains one of the most difficult challenges due to its inherent ill-posedness. While traditional numerical methods often struggle with the computational burden and regularization requirements of these problems, neural networks have recently emerged as a highly viable alternative, offering the ability to learn complex, non-linear mappings and provide efficient, real-time inference. This paper presents a framework for the MEG forward and inverse problems, integrating finite element modeling with neural network techniques. The forward problem is solved using FEniCS to model the electric potential governed by the Poisson equation on a realistic anatomical brain mesh, with magnetic fields computed via the Biot-Savart law. For the inverse problem, we introduce a Physics-Informed Neural Network (PINN) approach in order to deal with the ill condition of the problem. Unlike purely data-driven deep learning approaches that treat this problem as a black box learned from massive datasets, the proposed PINN framework directly embeds the governing physics--Maxwells equations and the Biot-Savart law--into the loss function, ensuring that the reconstructed sources satisfy the fundamental electromagnetic laws even in data-scarce regimes. We validate the framework on a high-resolution anatomical mesh and compared against the standard Minimum Norm Estimation (MNE). Results demonstrate that the PINN approach achieves a 30.2% improvement over the MNE baseline.

Electroencephalography (EEG) source imaging remains difficult to compare systematically because inverse solvers are distributed across different software packages, programming languages, and evaluation protocols. We present a frozen four-scenario EEG benchmark of 106 solvers evaluated on a shared BioSemi-32 / ico3 setup, together with invertmeeg, an open-source Python package that currently exposes 118 inverse solvers through a consistent two-step interface built on the MNE-Python ecosystem. The benchmark spans focal, multi-source, spatially extended, and low-SNR source configurations and uses earth movers distance (EMD) as the primary metric, with average precision (AP), mean localization error (MLE), and correlation used for complementary ranking. Across this benchmark, no single solver dominates every regime: flexible subspace and hybrid methods perform best overall, while Bayesian methods remain particularly competitive under extended-source and low-SNR conditions. The package is available via pip install invertmeeg and imported in Python as invert.

Longitudinal probing of structural connectivity via diffusion magnetic resonance imaging (dMRI) is experiencing uptake. However, the detection of biological effects is significantly hampered by the limitations of cross-sectional streamline tractography, where even small changes in the dMRI signal can produce drastically different trajectories and therefore quantitative parameterisation; if not properly dealt with, such effects will manifest as spurious longitudinal change, which can obscure subtle biological differences. To overcome this challenge, we here introduce a novel quantitative streamline tractography framework tailored for longitudinal analysis, wherein an individuals streamline trajectories remain fixed throughout the analysis, allowing only their ascribed density weights to vary between sessions. We present two strategies by which these quantitative streamline weights can be determined, both extensions of the widely adopted SIFT2 method. The performance of this framework is benchmarked against cross-sectional reconstruction with and without SIFT2 optimisation, in both in silico dMRI phantoms with known ground truths and three distinct human in vivo cohorts with clear a priori expectations of biological effects. We demonstrate that the proposed framework drastically reduces methodological imprecisions in synthetic dMRI phantoms and enhances statistical sensitivity and specificity to biological effects in human cohorts, enabling robust longitudinal quantification of structural connectivity.

Multi-site diffusion MRI studies face scanner-induced variability that can obscure biological signal. Harmonization methods such as ComBat have been developed to address this, but have been evaluated primarily on diffusion scalar metrics. Whether scanner reproducibility differs across fundamentally distinct tract-derived representations has not been systematically compared. Here, we compared the reproducibility of three metric families (diffusion, shape, and connectivity) across 36 association tracts using the MASiVar dataset (5 subjects, 4 scanners, 27 sessions). We assessed intraclass correlation coefficients (ICC) and multivariate subject discrimination at baseline, under dimensionality reduction, and after ComBat harmonization. At baseline, shape metrics showed the highest reproducibility (median ICC 0.69), followed by connectivity (0.49) and diffusion (0.34). Shape and connectivity achieved comparable subject discrimination (both 1.75), significantly exceeding diffusion (1.23). ComBat harmonization improved all families but harmonized diffusion (0.58) remained below unharmonized shape (0.69), indicating that metric family selection remains consequential even after harmonization. Under low-dimensional representation, connectivity showed the largest gains (ICC 0.86, subject discrimination 3.0), exceeding other families at any dimensionality. Analysis of principal component loadings identified a small number of cortical regions per tract (median 6) that capture 95% of the reproducible connectivity signal, providing a per-tract reference for selecting the most informative regions in future multi-site studies. These findings indicate that the choice of which tract-derived metrics to analyze in multi-site studies deserves at least as much consideration as how to harmonize them.

In the rapidly growing field of dynamic connectivity (dyC)-based real-time functional magnetic resonance imaging (rtfMRI) neurofeedback training, participants receive feedback on the dynamical coupling of brain nodes based on real-time dynamic connectivity measurements between multiple regions-of-interest. Given the fundamental role of regions-of-interest in this context, their signal-to-noise-ratio is critical. In closely related activity-based rtfMRI neurofeedback, region-of-interest selection is guided by functional localization methods informed by the BOLD-signal during a functional localizer experiment conducted prior to neurofeedback. However, methods that account for dynamic coupling of multiple brain regions remain lacking. Here, we develop a method based on dynamic connectivity between two brain areas that aligns with conventional activity-based localization approaches and is applicable for rtfMRI-NF. The proposed pipeline adapts processing steps from activity-based methods and was successfully implemented in a dyC-informed neurofeedback study. We observed a significant increase in dyC signal-to-noise-ratio during the localizer experiment and the first three blocks of neurofeedback training. Importantly, dyC values also correlated with anxiodepressive symptom levels, serving as proof of sensitivity, as changes in dyC reflect not only task-related neural dynamics but also individual differences in anxiodepressive traits. Our findings demonstrate that functional localization methods can be extended to dynamic connectivity, improving rtfMRI feedback accuracy, and with evidence to serve as sensitive measure for emotional states. This approach may enable precise targeting of coupled brain regions in neurofeedback and holds potential for personalized clinical applications, with broader implications beyond neurofeedback discussed.

The complex interplay between brain regions that support emotional experience and their link to individual differences is a topic of active research. Additionally, there has been growing interest in using functional gradients to investigate human cortical organization during both rest and film fMRI. Among these, several studies demonstrated improved brain fingerprinting performance, reflecting greater neural identification capability of film fMRI against rest fMRI despite higher subject synchronization during film-watching than in rest. Comparably, in this work we study the relation between individual differences, in particular, state anxiety and openness scores, and brain activity during the processing of various emotional scenes in films, through functional gradients. Next to including subcortical areas, we also propose a new approach of computing functional gradients based on a subset of frames selected using emotional annotation data of films, resulting in emotion-informed functional gradients. Then we evaluate the variance in emotion-informed gradients across subjects and employ these same gradients in the prediction of individual differences. For emotion-informed functional gradients, the highest predictability of state anxiety was found for scenes of negative valence and medium-high arousal, corresponding to the typical location of anxiety within the valence-arousal-power emotional space. Additionally, predictability of state anxiety was negatively correlated to inter-subject variability. In contrast, predictability of openness was found to be highest during scenes with low arousal and positively correlated to inter-subject variability. In essence, our results first show that macroscale brain organization is affected by emotional experience, and that frame selection based on the latter can be useful to remove non-subject-specific variability while extracting subject-specific information related to the emotion experience. It also demonstrates that frame selection increases inter-subject variability allowing the extraction of more subject-specific information. Thus, expanding on the idea of brain fingerprint in film fMRI, we argue that emotional experiences enhance disentanglement of various domain of individual differences. Moreover, depending on the individual difference of interest, fMRI acquired during more or less constrained paradigms would be more suitable to reveal different properties of brain function.

Dynamic positron emission tomography (PET) provides a powerful tool for studying in vivo neurochemical processes, including transient neurotransmitter release. However, widely used models such as the simplified reference tissue model (SRTM) assume time-invariant kinetic parameters, limiting their ability to capture dynamic changes in binding. Existing extensions introduce time-varying effects through parametric response functions or basis expansions, but are often constrained by restrictive functional assumptions, computational complexity, or limited uncertainty quantification. We propose a latent Gaussian process extension of SRTM (LGPE-SRTM), in which the apparent efflux parameter is modeled as a smooth, time-varying function within a hierarchical framework. By applying a first-order implicit discretization of the governing differential equation, the model admits a representation that is linear in all kinetic parameters in the mean structure, while nonlinearity is confined to a parameter-dependent covariance. This yields a conditionally linear mixed-effects model with structured covariance, enabling efficient likelihood-based inference. The proposed approach integrates mechanistic modeling, nonparametric flexibility, and hierarchical inference in a unified and computationally scalable framework. By representing functional effects on a shared temporal domain, all core computations reduce to operations on low-dimensional matrices whose size is independent of the number of subjects. This enables robust population-level inference on timevarying neurotransmitter dynamics without imposing restrictive parametric forms. The method is demonstrated on both simulated and empirical PET data, where it accurately recovers transient effects, provides well-calibrated uncertainty, and distinguishes constant from time-varying dynamics.

The MNI152 template is widely regarded as a representative average brain in neuroimaging, computational modeling, and neuromodulation research; however, its fidelity as a true population mean has not been systematically evaluated. In this study, we compared the MNI152 6th generation nonlinear template to 436 individual MRI from a publicly available dataset, including Asian, Black, and White participants. We quantified the gross brain dimensions and extracted the mean scaling, shear, and voxel-wise Jacobian determinants from the linear and nonlinear registrations between the template and each subject. Across all racial groups, the MNI152 brain exhibited substantially larger radii than the population means, with z-scores frequently exceeding 1.0. The linear scaling factors indicated consistent contraction of the template, and voxel-wise Jacobian fields revealed spatially heterogeneous deformations, demonstrating that the template differs from real brains in both size and shape. These findings suggest that the MNI152 template does not reflect the average morphology of contemporary population samples and that linear registration alone cannot resolve these discrepancies. Therefore, more robust and unbiased template-generation pipelines may be necessary for applications requiring anatomically accurate head models.

The rapid evolution of machine learning (ML) methods has yielded promising results in human brain neuroscience. However, the reproducibility of ML applications in neuroimaging remains limited, challenging the generalizability of inferences to broader populations. In addition to the inherent variability of the brain activity (both in healthy and pathological states), poor reproducibility is further enhanced by inconsistencies in data preprocessing techniques and methods for calculating functional connectivity (FC), which are used as parameters for brain state classification. To systematically assess the impact of abovementioned factors on ML applications to fMRI data, we benchmarked a comprehensive set of FC analysis pipelines for the classification task between fMRI data recorded in two fundamentally different states: eyes open and eyes closed. In contrast to studies involving heterogeneous clinical populations or using complex cognitive tasks, our controlled experimental design - based on two independent datasets of healthy participants collected in different laboratories - minimizes variability related to a task design or pathological brain states. Classification accuracy and reproducibility were compared for 256 distinct FC analysis pipelines, covering common preprocessing approaches, brain parcellation schemes, and connectivity metrics. Notably, we employed two ways of validation: a direct cross-site validation strategy - when a model was trained on one site and tested on another, and few-shot domain adaptation - when a few samples of testing site were added to the train set. Despite the substantial variability in pipeline configurations, we observed consistently high classification accuracy ([~]80%), confirming that FC-based models can robustly discriminate between well-defined brain states (eye conditions) across different acquisition sites. Best results both in terms of classification accuracy and stability were observed using Pearson correlation and tangent space parametrization as FC, Brainnetome as atlas, and confound regression strategies based on the CompCor method. These findings highlight the resilience of rs-fMRI FC-derived characteristics to methodological variation and support their utility in the discovery of biomarkers, particularly in settings that involve stable and reproducible brain states.

AO_SCPLOWBSTRACTC_SCPLOWCerebrospinal and interstitial fluid transport play a central role in brain metabolic waste clearance, yet non-invasive quantification of deep-brain flow dynamics remains challenging. Magnetic Resonance Artificial Intelligence Velocimetry (MR-AIV) is a physics-informed neural network framework that infers three-dimensional velocity, pressure, and permeability fields from dynamic contrastenhanced MRI by embedding porous-media flow physics into the learning process. Here, we present a methodological refinement and systematic evaluation of MR-AIV. We introduce a universal, anatomically informed, region-of-interest-based permeability initialization that improves anatomical alignment and physical consistency across subjects. We quantify the sensitivity of inferred fields to key modelling choices, including initialization strategies, permeability bounds, diffusivity assumptions, signal-concentration relationships, and measurement noise. Across these conditions, MR-AIV yields stable velocity and permeability estimates with preserved spatial structure. Together, these results establish practical guidelines and identify stable operating regimes for reliable deployment of MR-AIV. By improving robustness and reproducibility, this work strengthens MR-AIV as a minimally invasive approach for mapping brain-wide porous fluid transport and supports its application to studies of neurological health and disease.

Diffusion-weighted magnetic resonance imaging (dMRI) is used to study white matter microstructure and to delineate pathways by estimating fiber orientation distributions (FODs). Symmetric FODs represent the conventional model assuming antipodal symmetry in water diffusion. However, in complex regions with bending, branching or fanning fibers, this assumption is not guaranteed. To better capture such underlying fibers geometries, asymmetric FODs (A-FODs), derived from neighboring FODs, have been introduced. Here, we propose an Encoder-based Curvature-Aware Regularization (EnCAR) method for estimating A-FODs. Incorporating curvature features into the regularization weight applied to neighboring voxels improves reconstruction of A-FODs. A self-supervised Transformer network, combined with a Spherical Harmonics Semantic Encoder, learns region-specific regularization parameters from this local neighborhood to capture the diversity of fiber geometries across the brain. The EnCAR method was verified on the DiSCo challenge phantom, and applied to in vivo multi-shell Human data. The model estimated sharp, high-angular-resolution A-FODs that were well aligned with local fiber pathway. Compared with established FOD and A-FOD methods, it performed on par in regions dominated by symmetric FODs and outperformed them in complex asymmetric regions. Quantitative evaluation using the Asymmetry Index (ASI) and Model Discrepancy Index (MDI) confirmed improved consistency with the underlying diffusion signals. By ensuring smooth directional transitions, this work enhances the visibility of continuous fiber segments.

Functional brain mapping is an important tool to understand the organization of the human brain, both at the group level, but also to an increasing degree at the level of the individual. There are currently two main approaches to do so. Resting-state fMRI relies on inter-regional correlations of random fluctuations of the signal. In contrast, task-based localizers typically use a single-contrast between a task of interest and a matched control task to identify the location of a functional region in an individual brain. In this paper, we propose and evaluate a third approach: the use of multi-task batteries for both localization of a single functional region and parcellation of multiple functional regions. We show that multi-task localizers produce more consistent estimation of a single functional region across subjects than the single-contrast approach using the same amount of fMRI data. Furthermore, we demonstrate that the multi-task approach is sensitive to true inter-individual differences in region size, and does not suffer the same influence of signal-to-noise ratio that biases the single-contrast localizer. We then address the question of how to select tasks for the battery, and present a data-driven strategy that optimizes the characterization of a brain structure of interest. We show that such batteries outperform randomly selected batteries both for building individual parcellations as well as individual connectivity models. Finally, we demonstrate that an interspersed design - where all tasks are presented in each imaging run - yields more reliable results than splitting the tasks across different runs. We present an open source toolbox for the implementation of multi-task batteries, along with a library containing group-averaged activity patterns that can be used to optimize battery selection for different brain structures of interest.

Compared with model-based and phenomenological descriptions of the spatiotemporal dynamics of resting-brain activity, statistical characterizations of resting-state fMRI (rs-fMRI) data remain relatively underexplored. Some sophisticated analysis techniques, such as Mapper-based topological data analysis (TDA) and innovation-driven coactivation pattern analysis (iCAP), can distinguish real data from phase-randomized (PR) surrogates, suggesting that rs-fMRI data are not as simple as stationary Gaussian processes. However, the exact statistical properties that distinguish real rs-fMRI data from PR surrogates have not yet been determined. In this study, we conducted system identification analysis and surrogate data analysis to specify key statistical properties that allow TDA and iCAP to discriminate real rs-fMRI data from PR surrogates. We first analyzed rs-fMRI data concatenated across scans using autoregressive (AR) modeling and found that the scan-concatenated rs-fMRI data were weakly non-Gaussian. However, non-Gaussianity alone was insufficient to reproduce realistic TDA and iCAP results because of non-stationarity across scans. AR modeling of single-scan data revealed that rs-fMRI data were statistically indistinguishable from a Gaussian distribution within a single scan, although TDA and iCAP results still differed between the real data and PR surrogates. A new surrogate dataset designed to preserve non-stationarity successfully reproduced realistic TDA and iCAP results, suggesting that TDA and iCAP likely capture the non-stationarity of rs-fMRI data to distinguish it from PR surrogates. Together, these results indicate approximate Gaussianity and non-stationarity in rs-fMRI data, providing a data-driven and statistical characterization of resting-state brain activity that can serve as a quantitative reference for whole brain simulations and generative models.

In this paper, we propose a novel method that estimates time-resolved communities, or networks, from parcellated fMRI data based on instantaneous phase of the parcel timeseries. Importantly, each community label reference a limited phase range across time and subjects. It thus avoids the relabelling problem that is common for community detection algorithms. Our aim was to enhance the temporal resolution of brain network analysis in a whole-brain context. The method provides insights into how brain regions synchronize both within and across subjects and reorganize during task performance. It offers a complementary perspective on network integration and segregation as concurrent processes, quantified through differences in instantaneous phase. We employed HCP motor task fMRI data to exemplify its practical application. Prior to calculation of instantaneous phase, signal timeseries were decomposed into two signal modes with minimally overlapping frequency ranges. We show that task specific motor movements (hand, feet, tongue) can be separated from block-design related activation (visual and attention networks) where the former was found in the slower mode and the latter in the higher frequency mode.