Abstract Individual differences in the volumes of brain structures are often linked to various conditions, including Alzheimer’s disease, schizophrenia, and overall brain health. However, it remains unclear to what extent these differences reflect individual levels present from young adulthood or diverging aging trajectories from later ages. In this study, we analyze the aging dynamics of the volumes of six brain structures based on magnetic resonance imaging (MRI) scans from a large cross-cohort longitudinal sample of cognitively healthy adults (n = 8,311 with 18,520 MRIs, ages from 18 to 97 years). From general assumptions about structural brain dynamics and measurement noise, a stochastic dynamical model was fitted to the data to estimate both the variability and persistence of structural changes across adulthood. Using this model, we calculated how much of the variance of volumetric differences between individuals can be attributed to stable levels from young adulthood versus systematic changes at older ages, as well as the theoretical sensitivity of longitudinal studies to detect individual differences in change. The findings were as follows: 1) Before age 60 years, inter-individual differences in neuroanatomical volumes almost exclusively reflect stable differences between individuals, while the influence from systematic differences in rate-of-change increases thereafter; up to 50 % of the variation being due to differences in change at 80 years. In contrast, ventricular volume reflects differences in change from early adulthood. 2) Current brain-age models are unlikely to be sensitive to detect differences in aging trajectories. 3) Imaging studies have low reliability in detecting inter-individual brain changes before age 60. After 60 years, the study reliability increases sharply with longer intervals between scans and more modestly with additional intermediate observations. In conclusion, our results reinforce the view that it is critical to distinguish stable early-adulthood levels from systematic differences in change when studying adult brain aging.

Abstract Repetitive transcranial magnetic stimulation (rTMS) has been widely used in both clinical and research settings to modulate brain activity and assess brain plasticity. The effects of rTMS are generally believed to involve modulation of cortical excitability; protocols such as 10 Hz and iTBS increase cortical excitability, while 1Hz and cTBS are considered inhibitory. However, recent meta-analyses and studies have shown a wide array of interindividual responses to each protocol and poor reliability of the results. Here, we applied two sessions of the four most commonly used rTMS protocols (iTBS, cTBS, 10Hz, 1Hz) and Sham to the primary motor cortex and assessed their neurophysiological and behavioral effects. We hypothesized that relative to Sham, 10Hz and iTBS would increase motor excitability and modulate motor response times, while 1Hz and cTBS would have opposite effects. We recorded single-pulse TMS-EEG-EMG pre-rTMS and five and 25 minutes post-rTMS. Participants performed a sequential finger-tapping task (SFTT) 20 minutes pre-rTMS and 15 minutes post-rTMS. Our main outcome measures were: P30 amplitude, early LMFP (15-75 ms) area under-the-curve, MEP and SFTT completion time. We calculated the Intra-Class Correlation between each session for each measure. For each protocol, we found no consistent protocol-dependent group effect on MEPs, TEPs or behavior, with high intra-individual variability across sessions. An increase in SFTT completion time occurred for all active protocols but also Sham. The test-retest reliability of rTMS effects was low (ICC<0.5) for almost all protocols and metrics. This study is the first comprehensive assessment of the effects and reliability of the most commonly used rTMS protocols on cortical excitability and behavior. Our results suggest that rTMS effects are not reliable, at least at the single-session level, and that the inclusion of Sham rTMS is essential for identifying protocol-specific effects. These results also suggest that a fundamental and unbiased re-evaluation of potential rTMS mechanisms is necessary.

Sub-millimeter resolution functional magnetic resonance imaging (fMRI) at ultra-high field (≥ 7T) has offered an unprecedented opportunity to probe mesoscopic computations at a columnar or laminar level. However, its application has been primarily restricted to the neocortex. Inferior brain regions, particularly the hippocampus (HC), are challenging targets for laminar fMRI. Recent developments in acquisition methods have shown the feasibility of laminar recordings in the HC using gradient-echo blood oxygenation level-dependent (BOLD) contrast. Nonetheless, the spatial specificity of the BOLD signal is compromised by the draining veins’ bias. Cerebral blood volume (CBV)-sensitive sequences including vascular space occupancy (VASO) have emerged as a promising approach to capture the laminar activity with mitigated venous bias. Yet, its feasibility in the HC is unclear and challenged by methodological constraints. Here, we optimized VASO to mitigate the macrovasculature contribution in HC. By evaluating a series of advanced acquisition strategies tailored to HC, we obtained improved VASO signal quality with minimal artifacts. The optimized protocol was further validated with an autobiographical memory task. Our findings show that combining the high detection power of gradient-echo BOLD with the vein-bias-mitigated VASO contrast allows for differentiation between neural activity-related BOLD signals and those biased by draining veins. These results demonstrate the feasibility of submillimeter VASO acquired with conventional 7T scanners in the HC to map the circuit-level mechanisms of memory retrieval across HC subfields, laying a foundation to investigate the microcircuitry of HC-driven complex cognitive functions and their alterations in neurodegeneration and epilepsy.

Fixel-based analysis (FBA) has gained substantial interest for its ability to probe fibre-specific changes in the brain’s white matter from diffusion-weighted imaging data. However, the reproducibility and reliability of fixel-based measures across different scanners remains largely unknown. In this work, we present TRAMFIX (TRavelling Across Melbourne for FIXel-based analysis): a multisite dataset of traveling participants (n = 10 healthy adults) scanned across four 3T MRI scanners using a harmonised multi-shell diffusion-weighted imaging (DWI) protocol. DWI data were processed using two pipelines that can be adopted when performing multi-site FBA studies (site-specific vs. pooled processing). We extracted fixel-based measures of fibre density (FD), fibre cross-section (FC), and fibre density and cross-section (FDC) from the harmonized protocol. While the primary goal was to assess reproducibility and reliability of FBA metrics, we additionally computed diffusion tensor imaging (DTI)-based fractional anisotropy (FA) and mean diffusivity (MD) for the purposes of comparison with previous studies. Within-subject coefficients of variation (CVws) and intraclass correlation coefficients (ICC) of FBA and DTI measures were computed at multiple resolutions of computation and analysis: (i) the whole-brain averaged level, (ii) tract-level, and (iii) fixel- or voxel-level. Fixel-based metrics demonstrated high reproducibility and reliability at the whole-brain level (CVws ranging between 0.51% to 1.57% and ICC between 0.782 and 0.994). While reproducibility and reliability remained high for tract-averaged FBA measures (particularly the FC and FDC metrics), some tracts exhibited lower ICC values < 0.8 for the FD measure. When examining fixel-level reliability and reproducibility, clear spatial patterns emerged, with lower ICC across subcortical and cerebellar regions, and higher CVws at the cortical boundaries. FBA metrics demonstrated comparable, if not slightly better, reliability than tensor-based metrics derived from a subset of the same data. Our findings provide support for the reproducibility and reliability of fixel-based measures, highlighting their potential for use in multi-site FBA studies. Future work examining protocol-related differences, as well as appropriate harmonization strategies when pooling data across sites and scanners, will be valuable. To facilitate this, we provide the TRAMFIX dataset as a resource for investigating reproducibility, reliability, and harmonization of fixel-based analysis measures.

Many people experience aversive hypersensitivity (discomfort/visual stress) to stimuli such as bright lights, striped patterns, strobing, motion, or complex visual scenes such as supermarkets. Such sensory hypersensitivity is often associated with one or more of a range of neurological, psychiatric, and neurodevelopmental conditions or neurodivergence. The cortical mechanisms of sensory hypersensitivity, and reasons why it occurs with such a range of conditions, remain unknown. For three decades, theories have focused on excitation/inhibition balance, where visual discomfort reflects over-excitation relative to inhibition. Visual gamma oscillations induced by viewing stripes are an accepted index of excitation/inhibition, and are successfully modeled by a cortical circuit. Visual gamma is, therefore, predicted to be altered in people with high visual discomfort. We tested this in two studies. The first used circular moving gratings to evoke visual gamma, alongside self-reported scales for sensory sensitivity and for discomfort induced by viewing images (N = 166). We found no correlation of subjective sensitivity or discomfort with gamma frequency or amplitude (all r<0.1), or with the modeled excitation/inhibition parameters. In study 2, we recruited two groups of participants with high and low sensitivity to visual stripes (N = 23,27), and induced gamma with gratings of four different spatial frequencies. We found no group differences in gamma frequency, amplitude or modeled parameters. We conclude that visual discomfort is not simply explained by higher excitation/inhibition ratio in the visual cortex, despite the dominance of this assumed explanation.

The development and maturation of multisensory processing and integration depend on the brain’s ability to detect patterns and adjust behaviour based on feedback. However, the developmental dynamics of these mechanisms are not yet fully understood. This study examines neural and behavioural differences in multisensory statistical and reinforcement learning processes by comparing adults and children, and further investigates these processes across two learning tasks of varying difficulty in children. Using a discriminative choice and a match recognition multisensory associative learning task with embedded statistical stimulus regularities, 28 adults (19.0–30.9 years) and 2 groups of children (N = 28 and N = 29, 8.5–12.8 years) learned novel, non-linguistic associations of audio–visual or tactile–visual stimuli. Behaviourally, accuracy increased and reaction time decreased over time in all groups, and these learning effects were stronger in adults than in children and for audio–visual than for tactile–visual learning. Computational modelling revealed that adults employed more advanced learning strategies, with greater sensitivity to value differences, and more deterministic choice patterns. Behaviourally, reaction times were slower for more surprising trials across all groups. Neuroimaging showed that statistical surprise modulated activity in the lateral prefrontal cortex, intraparietal sulcus, and anterior cingulate cortex, while reward prediction errors were linked to striatal, medial frontal, and hippocampal regions. Reward prediction errors were consistent between adults and children and between task variants, whereas neural surprise processing emerged only in the match recognition task and showed developmental differences, indicating immaturity in children’s ability to detect statistical regularities. These findings highlight the importance of task design when comparing learning differences across developmental groups and showcase how reinforcement and statistical learning can be studied in parallel.

The integration of neuroimaging and transcriptomics data has revolutionized our understanding of complex brain phenomena and holds promise for linking molecular level genetic markers to brain dynamics and function. However, the involved data differ strongly across biological scales and acquisition modalities, presenting significant preprocessing and analytical challenges. In this context, multi-modal brain data exploration software offers a low-threshold, holistic, and non-biased approach to hypothesis generation that is complementary to literature search. This review assesses current data exploration tools that combine transcriptomic and neuroimaging data, evaluating them based on species coverage, data modalities, data integration approaches, visualization capabilities, and user data integration. We highlight the potential of these tools by discussing their applicability for example research questions and identify yet unaddressed workflows and promising approaches to support imaging transcriptomics analyses. This comprehensive survey aims to inform researchers in their selection of suitable software for their specific research needs and to guide the development of more effective multi-modal brain data exploration tools.

Identifying the role of compartmentalized inflammation within the central nervous system (CNS) and its association with chronic active lesions is essential for advancing our overall understanding of disease progression in multiple sclerosis (MS). In this study, we explore whether the inflammatory activity is higher in white matter (WM) tracts disrupted by paramagnetic rim lesions (PRLs) and if inflammation in PRL-disrupted WM tracts is associated with disability in people with MS. Forty-four MS patients and 11 healthy controls were included. 18 kDa-translocator protein positron emission tomography (TSPO-PET) with the 11C-PK11195 radioligand was used to measure the neuroinflammatory activity. The Network Modification Tool was used to identify WM tracts disrupted by PRLs and non-PRLs that were delineated on MRI. The Expanded Disability Status Scale was used to measure disability. MS patients had higher inflammatory activity in whole-brain WM compared to healthy controls (p = 0.001). Compared to patients without PRLs, patients with PRLs exhibited higher levels of inflammatory activity in the WM tracts disrupted by any type of lesions (p = 0.02) or PRLs (p = 0.004). In patients with at least one PRL, inflammatory activity was higher in WM tracts highly disrupted by PRLs compared to WM tracts highly disrupted by non-PRLs (p = 0.009). Elevated inflammatory activity in highly disrupted WM tracts was associated with increased disability in patients with PRL (p = 0.03), but not in patients without PRL (p = 0.2). This study suggests that patients with PRLs may exhibit more diffuse WM inflammation in addition to higher inflammation along WM tracts disrupted by PRLs compared to non-PRLs, which could contribute to larger lesion volumes and faster disability progression. Imaging PRLs may serve to identify patients with both focal and diffuse inflammation, guiding therapeutic interventions aimed at reducing inflammation and preventing progressive disability in MS.

Microstructure.jl is a Julia package designed for probabilistic estimation of tissue microstructural parameters from diffusion or combined diffusion–relaxometry MRI data. It provides a flexible and extensible framework for defining compartment models and includes robust and unified estimators for parameter fitting and uncertainty quantification. The package incorporates several established models from the literature, such as the spherical mean technique (SMT), standard model imaging (SMI), and soma and neurite density imaging (SANDI), along with their extensions for analyzing combined diffusion and T2 mapping data acquired at multiple echo times. For parameter estimation, it features methods such as Markov Chain Monte Carlo (MCMC) sampling and Monte Carlo dropout with neural networks, which provide probabilistic estimates by approximating the posterior distributions of model parameters. In this study, we introduce the major modules, functionality, and design of this package. We demonstrate its usage in optimizing acquisition protocols and evaluating fitting performance with synthesized ex vivo datasets for estimating the axon diameter index. We also demonstrate its practical applications for parameter estimation and uncertainty quantification with publicly available in vivo datasets: (1) SMT on 3 b-shell and 1.25-mm isotropic resolution data from the human connectome project, (2) SANDI on 2.0-mm isotropic resolution microstructure data from the high-gradient Connectome 1.0 scanner, and (3) SMI with a free water compartment on 2 b-shell, multi-echo-time, 2.5-mm isotropic resolution data from a Prisma scanner. Microstructure.jl is applicable to in vivo and ex vivo imaging data acquired with typical research, high-performance, or pre-clinical scanners.

Normative modelling in neuroimaging provides a powerful framework for quantifying individual deviations from expected brain measures as a function of relevant covariates. While earlier methods focused on analysing distinct variables in isolation, an increasing number of deep learning-based approaches are emerging to handle multiple response variables simultaneously. Conditional variational autoencoders (cVAEs) have previously been applied in this context and show promise for multivariate modelling. However, existing inference methods still face challenges in providing reliable probabilistic predictions, limiting their effectiveness as true normative models. In this study, we propose an enhanced cVAE-based framework that generates predictions directly from covariates through prior-sampling inference. This approach aligns with normative modelling principles while leveraging deep learning to handle high-dimensional data. We demonstrate the effectiveness of our approach using 195 imaging-derived phenotypes (IDPs), including morphometric features (cortical thickness, cortical volume, subcortical volume) and white matter hyperintensity (WMH) volumes, as a test case. Our dataset includes 8,551 normotensive and 18,180 hypertensive participants from the UK Biobank. We benchmarked our model against three well-established normative modelling techniques, including Generalised Additive Models for Location, Scale, and Shape (GAMLSS), Multivariate Fractional Polynomial Regression (MFPR), and Hierarchical Bayesian Regression (HBR), as well as the conventional posterior-sampling cVAE approach employed in existing autoencoder-based normative models. Through comparative analyses, our results show that the proposed cVAE-based framework achieves performance comparable with well-established normative models across various metrics, while appropriately capturing individual deviations associated with hypertension severity. Furthermore, our inference strategy demonstrates superior covariate sensitivity compared with those used in existing cVAE-based normative models, with deviations derived from our method showing better sensitivity to hypertension severity. Beyond predictive performance, our study also offers a comprehensive correlational mapping of the relationships between hypertension and brain structural damage. Taken together, this work highlights the promise of deep learning-based normative modelling for complex datasets such as neuroimaging, paving the way for personalised brain health assessment and early detection of neurological disorders.