Imaging Neuroscience

Recent advances in fetal fMRI present a new opportunity for neuroscience to study functional human brain connectivity at the time of its emergence. Progress in the field, however, has been hampered by the lack of openly available datasets that can be exploited by researchers across disciplines to develop methods that would address the unique challenges associated with imaging and analysing functional brain in utero, such as unconstrained head motion, dynamically evolving geometric distortions, or inherently low signal-to-noise ratio. Here we describe the developing Human Connectome Project’s release of the largest open access fetal fMRI dataset to date, containing 275 scans from 255 foetuses and spanning the period of 20.86 to 38.29 post-menstrual weeks. We present a systematic approach to its pre-processing, implementing multi-band soft SENSE reconstruction, dynamic distortion corrections via phase unwrapping method, slice-to-volume reconstruction and a tailored temporal filtering model, with attention to the prominent sources of structured noise in the in utero fMRI. The dataset is accompanied with an advanced registration infrastructure, enabling group-level data fusion, and contains outputs from the main intermediate processing steps. This allows for various levels of data exploration by the imaging and neuroscientific community, starting from the development of robust pipelines for anatomical and temporal corrections to methods for elucidating the development of functional connectivity in utero. By providing a high-quality template for further method development and benchmarking, the release of the dataset will help to advance fetal fMRI to its deserved and timely place at the forefront of the efforts to build a life-long connectome of the human brain.

Recent findings challenge traditional views of the Default Mode Network (DMN) as purely task-negative or self-oriented, showing increased DMN activity during demanding switches between externally-focused tasks (Crittenden et al., 2015; Smith et al., 2018; A. X. Zhou et al., 2024). However, it is unclear what modulates the DMN at switches, with transitions within a stimulus domain activating DMN regions in some studies but not others. Differences in the number of tasks suggest that complexity or structure of the set of tasks may be important. In this fMRI study, we examined whether the DMN’s response to task switches depended on the number of tasks that could be encountered in a run, or on abstract task groupings defined by the temporal order in which they were learnt at instruction. Core DMN activation at task switches was unaffected by the number of currently relevant tasks. Instead, it depended on the order in which groups of tasks had been learnt. Multivariate decoding revealed that Core DMN hierarchically represented individual tasks, task domains, and higher-order task groupings based on instruction order. We suggest that, as the complexity of instructions increases, rules are increasingly organised into higher-level chunks, and Core DMN activity is the highest at switches between chunks.

The human brain undergoes substantial developmental changes in the first 5 years of life. Particularly in the white matter, myelination of axons occurs near birth and continues at a rapid pace during the first 2 to 3 years. Diffusion MRI (dMRI) has revolutionized our understanding of developmental trajectories in white matter. However, the mm-resolution of in vivo techniques bears significant limitation in revealing the microstructure of the developing brain. Polarization sensitive optical coherence tomography (PSOCT) is a three-dimensional (3D) optical imaging technique that uses polarized light interferometry to target myelinated fiber tracts with micrometer resolution. Previous studies have shown that PSOCT contributes significantly to the elucidation of myelin content and quantification of fiber orientation in adult human brains. However, the use of PSOCT in developing human brains has not been reported. In this study, we established the feasibility of using the PSOCT technique to reveal brain development during the first 5 years of life, compared with ex vivo dMRI. The results showed that the optical properties of PSOCT quantitatively reveal the myelination process in young children. The imaging contrast of the optic axis orientation is a sensitive measure of fiber orientations in largely unmyelinated brains as young as 3 months old. The micrometer resolution of PSOCT provides substantially enriched information about complex fiber networks and complements submillimeter dMRI. This new optical tool offers great potential to reveal the white matter structures in normal neurodevelopment and developmental disorders in unprecedented detail.

The mechanisms linking perceptual and memory representations in the brain are not yet fully understood. In the early visual cortex, perception and memory are known to share similar neural representations, but how they interact beyond early visual cortex is less clear. Recent work identified that scene-perception and scene-memory areas on the lateral and ventral surfaces of the brain are linked via a shared but opponent visuospatial coding scheme, with spatially specific visual responses in the absence of traditionally defined retinotopic maps. This shared visuospatial coding may provide a framework for perceptual-memory interactions. Here, we test whether the pattern of visuospatial coding within category-selective memory areas of the medial parietal cortex structures responses during memory recall and visual perception. Using functional magnetic resonance imaging, we observe signatures of visuospatial coding in the form of population receptive fields (pRFs) with both positive and negative response profiles within medial parietal cortex. Crucially, the more dissimilar the timeseries of a pair of positive/negative pRFs within a region, the more dissimilar their responses during both memory recall and visual perception. These are tasks that place very different demands on these regions: internally oriented memory recall versus externally oriented visual perception. These data extend recent work to suggest that the interplay between pRFs with opponent visuospatial coding may play a vital role in integrating information across different representational spaces.

Anatomic tracing is the gold standard tool for delineating brain connections and for validating more recently developed imaging approaches such as diffusion MRI tractography. A key step in the analysis of data from tracer experiments is the careful, manual charting of fiber trajectories on histological sections. This is a very time-consuming process, which limits the amount of annotated tracer data that are available for validation studies. Thus, there is a need to accelerate this process by developing a method for computer-assisted segmentation. Such a method must be robust to the common artifacts in tracer data, including variations in the intensity of stained axons and background, as well as spatial distortions introduced by sectioning and mounting the tissue. The method should also achieve satisfactory performance using limited manually charted data for training. Here, we propose the first deep-learning method, with a self-supervised loss function, for segmentation of fiber bundles on histological sections from macaque brains that have received tracer injections. We address the limited availability of manual labels with a semi-supervised training technique that takes advantage of unlabeled data to improve performance. We also introduce anatomic and across-section continuity constraints to improve accuracy. We show that our method can be trained on manually charted sections from a single case and segment unseen sections from different cases, with a true positive rate of ∼0.80. We further demonstrate the utility of our method by quantifying the density of fiber bundles as they travel through different white-matter pathways. We show that fiber bundles originating in the same injection site have different levels of density when they travel through different pathways, a finding that can have implications for microstructure-informed tractography methods. The code for our method is available at https://github.com/v-sundaresan/fiberbundle_seg_tracing.

It is a fundamental question in the spatial navigation literature how different spatial cues are unified to form a coherent spatial map of the space. Landmarks and self-motion cues are two major spatial cue types, which recruit relatively independent cognitive processes that dynamically interact with each other during navigation. In our previous studies, we developed two novel memory-dependent paradigms to contrast visual landmarks and visual self-motion cues in the desktop virtual reality environment. Participants visited the four test locations arranged evenly along a linear track in predetermined sequences. While at each test location, they performed a spatial judgment relying on memory. Using ultra-high field fMRI at 7 Tesla, we found that the human entorhinal cortex (EC) and retrosplenial cortex (RSC) exhibited cue-specific location-based spatial representations in the form of fMRI adaptation (fMRIa), meaning that the closer the two successively visited locations were to each other, the greater the suppression in the brain activation. In the current study, we re-analyzed the same fMRI datasets from our previous studies by performing the representational similarity analysis (RSA), an approach complementary to the fMRIa analysis in assessing neural representations. RSA’s rationale is that the closer two locations are to each other in the space, the more similar multi-voxel patterns of brain activation they should elicit. The results showed that RSC contained RSA-based neural representations of spatial locations for both landmarks and self-motion cues, which were overall driven by subjective response (participant’s self-reported location) instead of objective location (participant’s actual location). These representations were generalizable between the two cue types in terms of response, indicating cue-independent spatial representations. Combined with our previous finding of cue-specific fMRIa-based spatial representations in RSC, our study demonstrates the coexistence of cue-specific and cue-independent spatial representations in RSC. Our findings suggest that RSC plays a crucial role in unifying various spatial sensory inputs into coherent spatial representations, supporting memory-oriented navigation behavior.

Cerebrovascular reactivity (CVR), the ability of cerebral blood vessels to dilate or constrict in response to a vasoactive stimulus, is a clinically useful measure of cerebrovascular health. CVR is often measured using a breath-hold task to modulate blood CO2 levels during an fMRI scan. Measuring end-tidal CO2 (PETCO2) with a nasal cannula during the task allows CVR amplitude to be calculated in standard units (vascular response per unit change in CO2, or %BOLD/mmHg) and CVR delay to be calculated in seconds. The use of standard units allows for normative CVR ranges to be established and for CVR comparisons to be made across subjects and scan sessions. Although breath holding can be successfully performed by diverse patient populations, obtaining accurate PETCO2 measurements requires additional task compliance; specifically, participants must breathe exclusively through their nose and exhale immediately before and after each breath hold. Meeting these requirements is challenging, even in healthy participants, and this has limited the translational potential of breath-hold fMRI for CVR mapping. Previous work has focused on using alternative regressors such as respiration volume per time (RVT), derived from respiration-belt measurements, to map CVR. Because measuring RVT does not require additional task compliance from participants, it is a more feasible measure than PETCO2. However, using RVT does not produce CVR amplitude in standard units. In this work, we explored how to achieve CVR amplitude maps, in standard units, and CVR delay maps, when breath-hold task PETCO2 data quality is low. First, we evaluated whether RVT could be scaled to units of mmHg using a subset of PETCO2 data of sufficiently high quality. Second, we explored whether a PETCO2 timeseries predicted from RVT using deep learning allows for more accurate CVR measurements. Using a dense-mapping breath-hold fMRI dataset, we showed that both rescaled RVT and rescaled, predicted PETCO2 can be used to produce maps of CVR amplitude in standard units and CVR delay with strong absolute agreement to ground-truth maps. The rescaled, predicted PETCO2 regressor resulted in superior accuracy for both CVR amplitude and delay. In an individual with regions of increased CVR delay due to Moyamoya disease, the predicted PETCO2 regressor also provided greater sensitivity to pathology than RVT. Ultimately, this work will increase the clinical applicability of CVR in populations exhibiting decreased task compliance.

State-of-the-art performance in electroencephalography (EEG) decoding tasks is currently often achieved with either Deep-Learning (DL) or Riemannian-Geometry-based decoders (RBDs). Recently, there is growing interest in Deep Riemannian Networks (DRNs) possibly combining the advantages of both previous classes of methods. However, there are still a range of topics where additional insight is needed to pave the way for a more widespread application of DRNs in EEG. These include architecture design questions such as network size and end-to-end ability. How these factors affect model performance has not been explored. Additionally, it is not clear how the data within these networks are transformed, and whether this would correlate with traditional EEG decoding. Our study aims to lay the groundwork in the area of these topics through the analysis of DRNs for EEG with a wide range of hyperparameters. Networks were tested on five public EEG datasets and compared with state-of-the-art ConvNets. Here, we propose end-to-end EEG SPDNet (EE(G)-SPDNet), and we show that this wide, end-to-end DRN can outperform the ConvNets, and in doing so use physiologically plausible frequency regions. We also show that the end-to-end approach learns more complex filters than traditional bandpass filters targeting the classical alpha, beta, and gamma frequency bands of the EEG, and that performance can benefit from channel-specific filtering approaches. Additionally, architectural analysis revealed areas for further improvement due to the possible under utilisation of Riemannian specific information throughout the network. Our study, thus, shows how to design and train DRNs to infer task-related information from the raw EEG without the need of handcrafted filterbanks and highlights the potential of end-to-end DRNs such as EE(G)-SPDNet for high-performance EEG decoding.

Optically-pumped magnetometers (OPMs) are compact and lightweight sensors that can measure magnetic fields generated by current flow in neuronal assemblies in the brain. Such sensors enable construction of magnetoencephalography (MEG) instrumentation, with significant advantages over conventional MEG devices including adaptability to head size, enhanced movement tolerance, lower complexity and improved data quality. However, realising the potential of OPMs depends on our ability to perform system calibration – which means finding sensor locations, orientations, and the relationship between the sensor output and magnetic field (termed sensor gain). Such calibration is complex in OPM-MEG since, for example, OPM placement can change from subject to subject (unlike in conventional MEG where sensor locations/orientations are fixed). Here, we present two methods for calibration, both based on generating well-characterised magnetic fields across a sensor array. Our first device (the HALO) is a head mounted system that generates dipole-like fields from a set of coils. Our second (the matrix coil (MC)) generates fields using coils embedded in the walls of a magnetically shielded room. Our results show that both methods offer an accurate means to calibrate an OPM array (e.g. sensor locations within 2 mm of the ground truth) and that the calibrations produced by the two methods agree strongly with each other: reconstructed positions, orientations and gains differ on average by 2.0 mm; 1.2° and 1.3% between HALO and MC. When applied to data from human MEG experiments, both methods offer improved signal-to-noise ratio after beamforming suggesting that they give calibration parameters closer to the ground truth than presumed physical sensor coordinates and orientations. Both techniques are practical and easy to integrate into real-world MEG applications. This advances the field significantly closer to the routine use of OPMs for MEG recording.

Transcranial Alternating Current Stimulation (tACS) appears to modulate neuronal oscillations at the frequency of stimulation. Longer periods of stimulation with tACS (10-40 minutes) have shown to produce persistent changes, especially in alpha power (~ 8-12 Hz), whereas the efficacy of shorter periods of tACS (1-8 seconds) is less known. Thus, we investigated whether short periods of tACS applied to the somatosensory cortex elicits changes in alpha power following stimulation. With this aim, during simultaneous acquisition of MEG, we administered tACS and control (no-tACS) on separate days. We applied short trains of stimulation for durations of 10 s and 30 s at an individually adapted stimulation frequency (ISF). Each stimulation-train was followed by a 15 s interval. We calculated power changes in the post-stimulation intervals, relative to a baseline period, and the resulting Δpower was used to statistically test the difference between tACS and control conditions. We found significant elevations in power at ISF following tACS compared to control. The extent of this effect spanned bilaterally over the somatosensory and frontal regions. While the observed increase in power was most prominent around ISF (i.e., in the alpha band), power modulations were also observed in the beta-band. When comparing the two stimulation durations, 10 s of tACS produced greater increases in power (at ISF) than 30 s of tACS. This study validates that 10 s of tACS produces robust elevations of power in the somatosensory cortex at ISF, thereby establishing its potential for use in future studies.

The human hippocampus has been extensively studied at the macroscale using functional magnetic resonance imaging (fMRI) but the underlying microcircuits at the mesoscale (i.e., at the level of layers) are largely uninvestigated in humans. We target two fundamental questions fundamental to hippocampal laminar fMRI: How does the venous bias affect the interpretation of hippocampal laminar responses, and is it possible to establish a benchmark laminar fMRI experiment which robustly elicits single-subject hippocampal activation utilizing the most widely applied GRE-BOLD contrast. We comprehensively characterized GRE-BOLD responses as well as T2*, tSNR and physiological noise as a function of cortical depth in individual subfields of the human hippocampus. Our results show that the vascular architecture differs between subfields leading to subfield-specific laminar biases of GRE-BOLD responses. Using an autobiographical memory paradigm, we robustly acquired depth-specific BOLD responses in hippocampal subfields. In the CA1 and Subiculum subregions, our results indicate a more pronounced trisynaptic path input rather than dominant direct inputs from entorhinal cortex during autobiographical memory retrieval. Our study provides unique insights into the hippocampus at the mesoscale level, and will help interpreting hippocampal laminar fMRI responses and allow researchers to test mechanistic hypotheses of hippocampal function.

Multiple Sclerosis (MS) is an autoimmune disease of the central nervous system (CNS), affecting 2.8 million people worldwide, that presents multiple features, one of which is demyelination. Although treatments exist to manage the condition, no cure has been found to stop the progression of neurodegeneration. To develop new treatments and investigate the multiple systems impacted by MS, new imaging technologies are needed at the preclinical stage. Functional ultrasound imaging (fUS) has recently been emerged as a robust method to measure brain cerebral blood volume (CBV) dynamics as an indirect indicator of neural activity. This study aimed to quantify the amplitude of alteration of evoked hemodynamic response in the somatosensory cortex, and its potential link with demyelination in a mouse model of CNS demyelination induced by cuprizone. We demonstrate that extended demyelination leads to an increased hemodynamic response in the primary sensory cortex, both spatially and temporally, aligning with fMRI findings in MS patients. Second, using descriptors of the evoked cortical hemodynamic response, we demonstrate that certain parameters (the number of active pixels and the rise time), correlate with the level of Myelin Basic Protein in the primary sensory cortex and the thalamus, when taken together. Interestingly, the increased CBV is not associated with demyelination but instead reflects the well-documented vascular alteration described in MS. Moreover, these changes were absent in the thalamus, and in focalized demyelinated lesions induced by lysolecithin injection, suggesting the involvement of specific cortical mechanisms driven by oligodendrocyte depletion. In conclusion, our study introduces a novel, non-invasive functional approach for investigating vascular dysfunction in the context of MS, addressing an important yet understudied aspect in both pre-clinical and clinical research.

Resting-state functional connectivity is a widely used approach to study the functional brain network organization during early brain development. However, the estimation of functional connectivity networks in individual infants has been rather elusive due to the unique challenges involved with functional magnetic resonance imaging (fMRI) data from young populations. Here, we use fMRI data from the developing Human Connectome Project (dHCP) database to characterize individual variability in a large cohort of term-born infants (N = 289) using a novel data-driven Bayesian framework. To enhance alignment across individuals, the analysis was conducted exclusively on the cortical surface, employing surface-based registration guided by age-matched neonatal atlases. Using 10 minutes of resting-state fMRI data, we successfully estimated subject-level maps for eight brain networks along with individual functional parcellation maps that revealed differences between subjects. We also found a significant relationship between age and mean connectivity strength in all brain regions, including previously unreported findings in higher-order networks. These results illustrate the advantages of surface-based methods and Bayesian statistical approaches in uncovering individual variability within very young populations.

Much of a child’s early learning takes place during social interactions with others. Neural synchrony, the temporal alignment of individuals’ functional brain activity, is a neural mechanism that may support successful interaction, but its biological origins and sensitivity to environmental factors remain unknown. This study measures neural coherence between 4- to 6-year-old children and their mothers using wearable functional near-infrared spectroscopy (“fNIRS”) in a collaborative problem-solving hyperscanning paradigm. Best practices in fNIRS data processing are incorporated to optimise coherence quantification and extricate environmental- and task-related effects. Results suggest physiological noise in the extracerebral layer artificially inflated coherence strength in both oxygenated (“HbO2”) and deoxygenated (“HbR”) haemoglobin. Coherence remained stronger during collaborative than during individual problem solving in both chromophores after physiological noise reduction. Phase-scrambled pseudodyad analyses supported the interpretation that coherence during collaboration relates to temporal dynamics of interaction rather than to task- or environmental-related components. Strength of HbO2 coherence was positively related to collaborative task performance and negatively related to background maternal stress. HbR coherence was also related to task performance and maternal stress but the direction of results were mixed. Overall, this study provides new insight into the nature of neural coherence between 4- to 6-year-old children and their mothers during collaborative play.

Altered iron levels, detected using iron-sensitive MRI techniques such as quantitative susceptibility mapping (QSM), are observed in neurological disorders and may play a crucial role in disease pathophysiology. However, brain iron changes occur slowly, even in neurological diseases, and can be influenced by physiological or environmental factors that are difficult to quantify in the research or clinical settings. Therefore, novel analysis methods are needed to improve sensitivity to disease-related iron changes beyond conventional region-based approaches. This study introduces IRONMAP, Iron Network Mapping and Analysis Protocol, which is a novel network-based analysis method to evaluate over-time changes in magnetic susceptibility. With this technique, we analyzed short-term (<1 year) longitudinal QSM data from a cohort of people with multiple sclerosis (pwMS) and healthy controls (HCs) and assessed disease-related network patterns, comparing the new approach to a conventional per-region rate-of-change method. IRONMAP revealed over-time, MS-related brain iron abnormalities that were undetectable using the rate-of-change approach. IRONMAP was applicable at the per-subject level, improving binary classification of pwMS vs HCs compared to rate-of-change data alone (areas under the curve: 0.773 vs 0.636, p = 0.024). Further analysis revealed that the observed IRONMAP-derived HC network structure closely aligned with simulated networks based on healthy aging-related susceptibility data, suggesting that disruptions in normal aging-related iron changes may contribute to the network differences seen in pwMS. IRONMAP is applicable to various neurological diseases, including Alzheimer’s disease and Parkinson’s disease, and can be used between any set of brain regions. Our proposed technique may allow for the study of brain iron abnormalities over shorter timeframes than previously possible.

Sharing neuroimaging data upon a direct personal request can be challenging both for researchers who request the data and for those who agree to share their data. Unlike sharing through repositories under standardized protocols and data use/sharing agreements, each party often needs to negotiate the terms of sharing and use of data case by case. This negotiation unfolds against a complex backdrop of ethical and regulatory requirements along with technical hurdles related to data transfer and management. These challenges can significantly delay the data-sharing process, and if not properly addressed, lead to potential tensions and disputes between sharing parties. This study aims to help researchers navigate these challenges by examining what to consider during the process of data sharing and by offering recommendations and practical tips. We first divided the process of sharing data upon a direct personal request into six stages: requesting data, reviewing the applicability of and requirements under relevant laws and regulations, negotiating terms for sharing and use of data, preparing and transferring data, managing and analyzing data, and sharing the outcome of secondary analysis of data. For each stage, we identified factors to consider through a review of ethical principles for human subject research; individual institutions’ and funding agencies’ policies; and applicable regulations in the U.S. and E.U. We then provide practical insights from a large-scale ongoing neuroimaging data-sharing project led by one of the authors as a case study. In this case study, PET/MRI data from a total of 782 subjects were collected through direct personal requests across seven sites in the USA, Canada, the UK, Denmark, Germany, and Austria. The case study also revealed that researchers should typically expect to spend an average of 8 months on data sharing efforts, with the timeline extending up to 24 months in some cases due to additional data requests or necessary corrections. The current state of data sharing via direct requests is far from ideal and presents significant challenges, particularly for early career scientists, who often have a limited time frame—typically 2 to 3 years—to work on a project. The best practices and practical tips offered in this study will help researchers streamline the process of sharing neuroimaging data while minimizing friction and frustrations.

It has been proposed that autistic perception may be marked by a reduced influence of temporal context. Following this theory, prior exposure to a stimulus should lead to a weaker or absent alteration of the behavioral and neural response to the stimulus in autism, compared with a typical population. To examine these hypotheses, we recruited two samples of human volunteers: a student sample (N = 26), which we used to establish our analysis pipeline, and an adolescent sample (N = 36), which consisted of a group of autistic (N = 18) and a group of non-autistic (N = 18) participants. All participants were presented with visual stimulus streams consisting of novel and familiar image pairs, while they attentively monitored each stream. We recorded task performance and used magnetoencephalography (MEG) to measure neural responses, and to compare the responses with familiar and novel images. We found behavioral facilitation as well as a reduction of event-related field (ERF) amplitude for familiar, compared with novel, images in both samples. Crucially, we found statistical evidence against between-group effects of familiarity on both behavioral and neural responses in the adolescent sample, suggesting that the influence of familiarity is comparable between autistic and non-autistic adolescents. These findings challenge the notion that perception in autism is marked by a reduced influence of prior exposure.

Pregnancy is a psycho-neuro-endocrinological transition phase presenting a window of vulnerability for mental health. Emotion regulation, a transdiagnostic factor for psychopathology, is influenced by estradiol across the menstrual cycle on the behavioral and neural level. Whether this is also the case in the antepartum period remains unknown. For the first time, behavioral and neural emotion regulation were investigated in healthy pregnant females with extremely high estradiol levels during the second trimester (N = 15) using a functional magnetic resonance imaging (fMRI) paradigm. Results were compared with naturally-cycling females with high (N = 16) and low estradiol levels (N = 16). Although pregnant females reported the lowest trait use of cognitive reappraisal, all participants successfully regulated their emotions by applying cognitive reappraisal in the scanner. During downregulation of negative emotions, all females had increased activity in the left middle frontal gyrus. Pregnant females showed no significant differences in functional connectivity (psychophysiological interaction, resting-state) related to emotion regulation compared to the nonpregnant groups. However, group differences emerged for amygdala activation. In pregnant females, increased amygdala activity predicted reduced regulation success and was positively associated with depression scores. This first fMRI study during pregnancy indicates that depression scores are reflected in heightened amygdala activity already observable in the antepartum period. Thus, through its association with reduced regulation success, increased amygdala activity suggests a neural risk marker for peripartum mental health. The findings highlight the importance of investigating neural and behavioral emotion regulation in the ante- and postpartum period, eventually allowing enhanced identification, prevention, and treatment of peripartum mental ill-health.

Positron emission tomography (PET) and single photon emission computed tomography (SPECT) are essential molecular imaging tools for the in vivo investigation of neurotransmission. Traditionally, PET and SPECT images are analysed in a univariate manner, testing for changes in radiotracer binding in regions or voxels of interest independently of each other. Over the past decade, there has been an increasing interest in the so-called molecular connectivity approach that captures relationships of molecular imaging measures in different brain regions. Targeting these inter-regional interactions within a neuroreceptor system may allow to better understand complex brain functions. In this article, we provide a comprehensive review of molecular connectivity studies in the field of neurotransmission. We examine the expanding use of molecular connectivity approaches, highlighting their applications, advantages over traditional methods, and contributions to advancing neuroscientific knowledge. A systematic search in three bibliographic databases MEDLINE, EMBASE and Scopus on July 14, 2023, was conducted. A second search was rerun on April 4, 2024. Molecular imaging studies examining functional interactions across brain regions were included based on predefined inclusion and exclusion criteria. Thirty-nine studies were included in the scoping review. Studies were categorised based on the primary neurotransmitter system being targeted: dopamine, serotonin, opioid, muscarinic, glutamate and synaptic density. The most investigated system was the dopaminergic and the most investigated disease was Parkinson’s disease (PD). This review highlighted the diverse applications and methodologies in molecular connectivity research, particularly for neurodegenerative diseases and psychiatric disorders. Molecular connectivity research offers significant advantages over traditional methods, providing deeper insights into brain function and disease mechanisms. As the field continues to evolve, embracing these advanced methodologies will be essential to understand the complexities of the human brain and improve the robustness and applicability of research findings in clinical settings.

Velocity-selective arterial spin labelling (VSASL) MRI is insensitive to prolonged arterial transit time. This is an advantage over other arterial spin labelling schemes, where long arterial transit times can lead to bias. Therefore, VSASL can be used with greater confidence to study perfusion in the presence of long arterial transit times, such as in the ageing brain, in vascular pathologies, and cancer, or where arterial transit time changes, such as during measurement of cerebrovascular reactivity (CVR). However, when calculating perfusion (cerebral blood flow, CBF, in the brain) from VSASL signal, it is assumed that a vascular crushing module, defining the duration of the bolus, is applied before the arrival of the trailing edge. The early arrival of the trailing edge of the labelled bolus of blood will cause an underestimation of perfusion. Here, we measure bolus duration in adult, healthy human brains, both at rest and during elevated CBF during CO2 breathing (5% inspired CO2). Grey matter bolus duration was of 2.20 ± 0.35 s/2.22 ± 0.53 s/2.05 ± 0.34 s (2/3/4 cm/s vcutoff) at rest, in close agreement with a prior investigation. However, we observed a significant decrease in bolus duration during hypercapnia, and a matched reduction in CVR above a labelling delay of approximately 1.2 s. The reduction in CVR and bolus duration was spatially heterogenous, with shorter hypercapnic bolus durations observed in the frontal lobe (1.31 ± 0.54 s) and temporal lobes (1.36 ± 0.24 s), compared to the occipital lobe (1.50 ± 0.26 s). We place these results in the context of recommendations from a recent consensus paper, which recommends imaging 1.4 s after the label, which could lead to CBF underestimation in conditions with fast flow or during CVR measurements. These results can be used to inform the experimental design of future VSASL studies, to avoid underestimating perfusion by imaging after the arrival of the trailing edge of the labelled bolus.

Artificial neural networks trained in the field of artificial intelligence (AI) have emerged as key tools to model brain processes, sparking the idea of aligning network representations with brain dynamics to enhance performance on AI tasks. While this concept has gained support in the visual domain, we investigate here the feasibility of creating auditory artificial neural models directly aligned with individual brain activity. This objective raises major computational challenges, as models have to be trained directly with brain data, which is typically collected at a much smaller scale than data used to train AI models. We aimed to answer two key questions: (1) Can brain alignment of auditory models lead to improved brain encoding for novel, previously unseen stimuli? (2) Can brain alignment lead to generalisable representations of auditory signals that are useful for solving a variety of complex auditory tasks? To answer these questions, we relied on two massive datasets: a deep phenotyping dataset from the Courtois neuronal modelling project, where six subjects watched four seasons (36 hours) of the Friends TV series in functional magnetic resonance imaging and the HEAR benchmark, a large battery of downstream auditory tasks. We fine-tuned SoundNet, a small pretrained convolutional neural network with ~2.5M parameters. Aligning SoundNet with brain data from three seasons of Friends led to substantial improvement in brain encoding in the fourth season, extending beyond auditory and visual cortices. We also observed consistent performance gains on the HEAR benchmark, particularly for tasks with limited training data, where brain-aligned models performed comparably to the best-performing models regardless of size. We finally compared individual and group models, finding that individual models often matched or outperformed group models in both brain encoding and downstream task performance, highlighting the data efficiency of fine-tuning with individual brain data. Our results demonstrate the feasibility of aligning artificial neural network representations with individual brain activity during auditory processing, and suggest that this alignment is particularly beneficial for tasks with limited training data. Future research is needed to establish whether larger models can achieve even better performance and whether the observed gains extend to other tasks, particularly in the context of few shot learning.

Traditional tissue classification approaches in vivo use voxel intensities from conventional clinical magnetic resonance (MR) images for segmentation, which does not incorporate information about specific aspects of microstructure. With the Clustering for Anatomical Quantification and Evaluation (CAQE) framework, quantitative MRI measures can be used to classify tissue based only on microstructural features with no spatial enforcement, and pathological changes in disease can be evaluated. In this study, maps of whole-brain myelin water fraction, microscopic fractional anisotropy and tissue heterogeneity were used to classify brain tissue in 25 healthy participants. CAQE was then applied to 25 participants with multiple sclerosis (MS), where tissue classifications indicated areas of increased demyelination and axonal injury in white matter compared to a healthy average tissue classification. Severity scores were derived from tissue classifications to quantify diffuse white matter damage, and correlated significantly with cognitive ability in MS. The CAQE framework can be adapted for other applications and extended to use different quantitative MRI measures.

Disruption of the balance between excitatory and inhibitory neurotransmission (E-I balance) is thought to underlie many neurodevelopmental disorders; however, its study is typically restricted to adults, animal models and the lab-bench. Neurophysiological oscillations in the gamma frequency band relate closely to E-I balance, and a new technology – OPM-MEG – offers the possibility to measure such signals across the lifespan. We used OPM-MEG to measure gamma oscillations induced by visual stimulation in 101 participants, aged 2-34 years. We demonstrate a significantly changing spectrum with age, with low amplitude broadband gamma oscillations in children and high amplitude band limited oscillations in adults. We used a canonical cortical microcircuit to model these signals, revealing a significant decrease in the ratio of excitatory to inhibitory signalling with age in the superficial pyramidal neurons of the visual cortex. Our findings detail the first MEG metrics of gamma oscillations and their underlying generators from toddlerhood, providing a benchmark against which future studies can contextualise.

In neuroimaging and functional Magnetic Resonance Imaging (fMRI), many derived data are made openly available in public databases. These can be re-used to increase sample sizes in studies and thus, improve robustness. In fMRI studies, raw data are first preprocessed using a given analysis pipeline to obtain subject-level contrast maps, which are then combined into a group analysis. Typically, the subject-level analysis pipeline is identical for all participants. However, derived data shared on public databases often come from different workflows, which can lead to different results. Here, we investigate how this analytical variability, if not accounted for, can induce false positive detections in mega-analyses combining subject-level contrast maps processed with different pipelines. We use the HCP multi-pipeline dataset, containing contrast maps for N=1,080 participants of the HCP Young-Adult dataset, whose raw data were processed and analyzed with 24 different pipelines. We performed between-groups analyses with contrast maps from different pipelines in each group and estimated the rates of pipeline-induced detections. We show that, if not accounted for, analytical variability can lead to inflated false positive rates in studies combining data from different pipelines.

Being able to aggregate results from many acceptable data analysis pipelines (multiverse analyses) is a desirable feature in almost all aspects of imaging neuroscience. This is because multiple noise sources may contaminate the acquired imaging data, and different pipelines will attenuate or remove those noise source effects differentially. Here, we used multiple preprocessing pipelines that are known to impact the final results and conclusions of Positron Emission Tomography (PET) neuroimaging studies significantly. We developed conceptual and practical tools for statistical analyses that aggregate pipeline results and a new sensitivity analysis testing for hypotheses across pipelines, such as “no effect across all pipelines” or “at least one pipeline with no effect”. The proposed framework is generic and can be applied to any multiverse scenario. Code to reproduce all analyses and figures is openly available, including a step-by-step tutorial, so other researchers can carry out their own multiverse analysis.