Explore the vast range of titles available.

We assessed communication skills of 48 children without cerebral palsy (CP) treated with therapeutic hypothermia (TH) for neonatal hypoxic-ischemic encephalopathy (HIE) (cases) compared to 42 controls at early school-age and examined their association with white matter diffusion properties in both groups and 18-month Bayley-III developmental assessments in cases. Parents completed a Children’s Communication Checklist (CCC-2) yielding a General Communication Composite (GCC), structural and pragmatic language scores and autistic-type behavior score. GCC ≤ 54 and thresholds of structural and pragmatic language score differences defined language impairment. Using tract-based spatial statistics (TBSS), fractional anisotropy (FA) was compared between 31 cases and 35 controls. Compared to controls, cases had lower GCC ( p  = 0.02), structural ( p  = 0.03) and pragmatic language score ( p  = 0.04) and higher language impairments ( p  = 0.03). GCC correlated with FA in the mid-body of the corpus callosum, the cingulum and the superior longitudinal fasciculus ( p  < 0.05) in cases. Bayley-III Language Composite correlated with GCC (r = 0.34, p  = 0.017), structural (r = 0.34, p  = 0.02) and pragmatic (r = 0.32, p  = 0.03) language scores and autistic-type behaviors (r = 0.36, p  = 0.01).

Children cooled for HIE and who did not develop cerebral palsy (CP) still underperform at early school age in motor and cognitive domains and have altered supra-tentorial brain volumes and white matter connectivity. We obtained T1-weighted and diffusion-weighted MRI, motor (MABC-2) and cognitive (WISC-IV) scores from children aged 6–8 years who were cooled for HIE secondary to perinatal asphyxia without CP (cases), and controls matched for age, sex, and socioeconomic status. In 35 case children, we measured cerebellar growth from infancy (age 4–15 days after birth) to childhood. In childhood, cerebellar volumes were measured in 26 cases and 23 controls. Diffusion properties (mean diffusivity, MD and fractional anisotropy, FA) were calculated in 24 cases and 19 controls, in 9 cerebellar regions. Cases with FSIQ ≤ 85 had reduced growth of cerebellar width compared to those with FSIQ > 85 ( p  = 0.0005). Regional cerebellar volumes were smaller in cases compared to controls ( p  < 0.05); these differences were not significant when normalised to total brain volume. There were no case–control differences in MD or FA. Interposed nucleus volume was more strongly associated with IQ in cases than in controls ( p  = 0.0196). Other associations with developmental outcome did not differ between cases and controls.

•We compared dimensionality reduction methods prior to clustering for dFC analysis.•We measured clustering performance in synthetic datasets with multiple subjects.•Deep clustering out-performed PCA, UMAP and raw k-means.•We demonstrated the effect of dimensionality reduction on results from real data. [Display omitted] Dynamic functional connectivity (dFC) analysis of resting-state fMRI data is commonly performed by calculating sliding-window correlations (SWC), followed by k-means clustering in order to assign each window to a given state. Studies using synthetic data have shown that k-means performance is highly dependent on sliding window parameters and signal-to-noise ratio. Additionally, sources of heterogeneity between subjects may affect the accuracy of group-level clustering, thus affecting measurements of dFC state temporal properties such as dwell time and fractional occupancy. This may result in spurious conclusions regarding differences between groups (e.g. when comparing a clinical population to healthy controls). Therefore, is it important to quantify the ability of k-means to estimate dFC state temporal properties when applied to cohorts of multiple subjects, and to explore ways in which clustering performance can be maximised. Here, we explore the use of dimensionality reduction methods prior to clustering in order to map high-dimensional data to a lower dimensional space, providing salient features to the subsequent clustering step. We assess the use of deep autoencoders for dimensionality reduction prior to applying k-means clustering to the encoded data. We compare this deep clustering method to dimensionality reduction using principle component analysis (PCA), uniform manifold approximation and projection (UMAP), as well as applying k-means to the original feature space using either L1 or L2 distance. We provide extensive quantitative evaluation of clustering performance using synthetic datasets, representing data from multiple heterogeneous subjects. In synthetic data we find that deep clustering gives the best performance, while other approaches are often insufficient to capture temporal properties of dFC states. We then demonstrate the application of each method to real-world data from human subjects and show that the choice of dimensionality reduction method has a significant effect on group-level measurements of state temporal properties.

In contrast to blood‐oxygenation level‐dependent (BOLD) functional MRI (fMRI), which relies on changes in blood flow and oxygenation levels to infer brain activity, diffusion fMRI (DfMRI) investigates brain dynamics by monitoring alterations in the apparent diffusion coefficient (ADC) of water. These ADC changes may arise from fluctuations in neuronal morphology, providing a distinctive perspective on neural activity. The potential of ADC as an fMRI contrast (ADC‐fMRI) lies in its capacity to reveal neural activity independently of neurovascular coupling, thus yielding complementary insights into brain function. To demonstrate the specificity and value of ADC‐fMRI, both ADC‐ and BOLD‐fMRI data were collected at 3 T in human subjects during visual stimulation and motor tasks. The first aim of this study was to identify an acquisition design for ADC that minimises BOLD contributions. By examining the timings in responses, we report that ADC 0/1 timeseries (acquired with b values of 0 and 1 ms/μm2 $$ {\\upmu \\mathrm{m}}^2 $$ ) exhibit residual vascular contamination, while ADC 0.2/1 timeseries (with b values of 0.2 and 1 ms/μm2 $$ {\\upmu \\mathrm{m}}^2 $$ ) show minimal BOLD influence and higher sensitivity to neuromorphological coupling. Second, a general linear model was employed to identify activation clusters for ADC 0.2/1 and BOLD, from which the average ADC and BOLD responses were calculated. The negative ADC response exhibited a significantly reduced delay relative to the task onset and offset as compared to BOLD. This early onset further supports the notion that ADC is sensitive to neuromorphological rather than neurovascular coupling. Remarkably, in the group‐level analysis, positive BOLD activation clusters were detected in the visual and motor cortices, while the negative ADC clusters mainly highlighted pathways in white matter connected to the motor cortex. In the averaged individual level analysis, negative ADC activation clusters were also present in the visual cortex. This finding confirmed the reliability of negative ADC as an indicator of brain function, even in regions with lower vascularisation such as white matter. Finally, we established that ADC‐fMRI time courses yield the expected functional organisation of the visual system, including both grey and white matter regions of interest. Functional connectivity matrices were used to perform hierarchical clustering of brain regions, where ADC‐fMRI successfully reproduced the expected structure of the dorsal and ventral visual pathways. This organisation was not replicated with the b = 0.2 ms/μm2 $$ {\\upmu \\mathrm{m}}^2 $$diffusion‐weighted time courses, which can be seen as a proxy for BOLD (via T2‐weighting). These findings underscore the robustness of ADC time courses in functional MRI studies, offering complementary insights into BOLD‐fMRI regarding brain function and connectivity patterns. This article validates ADC‐fMRI as a tool complementing BOLD‐fMRI, detecting neural activity via neuromorphological coupling. In experiments with motor and visual stimuli at 3 T, ADC‐fMRI was shown to be less prone to vascular contamination and to capture white matter activity.

Objective To investigate if an association exists between motion artefacts on brain MRI and comprehension, co-ordination, or hyperactivity scores in children aged 6–8 years, cooled for neonatal encephalopathy (cases) and controls. Methods Case children ( n  = 50) without cerebral palsy were matched with 43 controls for age, sex, and socioeconomic status. Children underwent T1-weighted (T1w), diffusion-weighted image (DWI) brain MRI and cognitive, behavioural, and motor skills assessment. Stepwise multivariable logistic regression assessed associations between unsuccessful MRI and comprehension (including Weschler Intelligence Scale for Children (WISC-IV) verbal comprehension, working memory, processing speed and full-scale IQ), co-ordination (including Movement Assessment Battery for Children (MABC-2) balance, manual dexterity, aiming and catching, and total scores) and hyperactivity (including Strengths and Difficulties Questionnaire (SDQ) hyperactivity and total difficulties scores). Results Cases had lower odds of completing both T1w and DWIs (OR: 0.31, 95% CI 0.11–0.89). After adjusting for case-status and sex, lower MABC-2 balance score predicted unsuccessful T1w MRI (OR: 0.81, 95% CI 0.67–0.97, p  = 0.022). Processing speed was negatively correlated with relative motion on DWI ( r  = −0.25, p  = 0.026) and SDQ total difficulties score was lower for children with successful MRIs ( p  = 0.049). Conclusions Motion artefacts on brain MRI in early school-age children are related to the developmental profile. Impact Children who had moderate/severe neonatal encephalopathy are less likely to have successful MRI scans than matched controls. Motion artefact on MRI is associated with lower MABC-2 balance scores in both children who received therapeutic hypothermia for neonatal encephalopathy and matched controls, after controlling for case-status and sex. Exclusion of children with motion artefacts on brain MRI can introduce sampling bias, which impacts the utility of neuroimaging to understand the brain–behaviour relationship in children with functional impairments.

Functional MRI (fMRI) using the blood-oxygen level dependent (BOLD) signal provides valuable insight into grey matter activity. However, uncertainty surrounds the white matter BOLD signal. Apparent diffusion coefficient (ADC) offers an alternative fMRI contrast sensitive to transient cellular deformations during neural activity, facilitating detection of both grey and white matter activity. Further, through minimising vascular contamination, ADC-fMRI has the potential to overcome the limited temporal specificity of the BOLD signal. However, the use of linear diffusion encoding introduces sensitivity to fibre directionality, while averaging over multiple directions comes at great cost to temporal resolution. In this study, we used spherical b-tensor encoding to impart diffusion sensitisation in all directions per shot, providing an ADC-fMRI contrast capable of detecting activity independently of fibre directionality. We provide evidence from two task-based experiments on a clinical scanner that isotropic ADC-fMRI is more temporally specific than BOLD-fMRI, and offers more balanced mapping of grey and white matter activity. We further demonstrate that isotropic ADC-fMRI detects white matter activity independently of fibre direction, while linear ADC-fMRI preferentially detects activity in voxels containing fibres perpendicular to the diffusion encoding direction. Thus, isotropic ADC-fMRI opens avenues for investigation into whole-brain grey and white matter functional connectivity. Detecting neural signals in white matter remains a challenge. Here, the authors introduce an isotropic apparent diffusion coefficient fMRI contrast based on neuromorphological coupling, which provides sensitivity to neural signals in both grey and white matter in the human brain.

Resting-state functional magnetic resonance imaging (fMRI) is used to derive functional connectivity (FC) between brain regions. Typically, blood oxygen level-dependent (BOLD) contrast is used. However, BOLD’s reliance on neurovascular coupling poses challenges in reflecting brain activity accurately, leading to reduced sensitivity in white matter (WM). WM BOLD signals have long been considered physiological noise, although recent evidence shows that both stimulus-evoked and resting-state WM BOLD signals resemble those in gray matter (GM), albeit smaller in amplitude. We introduce apparent diffusion coefficient fMRI (ADC-fMRI) as a promising functional contrast for GM and WM FC, capturing activity-driven neuromorphological fluctuations. Our study compares BOLD-fMRI and ADC-fMRI FC in GM and WM, showing that ADC-fMRI mirrors BOLD-fMRI connectivity in GM, while capturing more robust FC in WM. ADC-fMRI displays higher average clustering and average node strength in WM, and higher inter-subject similarity, compared to BOLD. Taken together, this suggests that ADC-fMRI is a reliable tool for exploring FC that incorporates gray and white matter nodes in a novel way. de Riedmatten et al. propose apparent diffusion coefficient functional MRI as a novel contrast for assessing functional connectivity in both grey matter and white matter, addressing the limitations of blood oxygen level-dependent fMRI in capturing white matter activity. Their study demonstrates that this novel methods provides a more reliable white matter connectivity measure, with higher inter-subject similarity and reduced sensitivity to vascular noise, making it a promising alternative for investigating whole-brain functional connectivity.

Objective Neonatal imaging studies report corpus callosum abnormalities after neonatal hypoxic–ischaemic encephalopathy (HIE), but corpus callosum development and relation to cognition in childhood are unknown. Using magnetic resonance imaging (MRI), we examined the relationship between corpus callosum size, microstructure and cognitive and motor outcomes at early school‐age children cooled for HIE (cases) without cerebral palsy compared to healthy, matched controls. A secondary aim was to examine the impact of HIE‐related neonatal brain injury on corpus callosum size, microstructure and growth. Methods Participants aged 6–8 years underwent MRI, the Movement Assessment Battery for Children Second Edition and Wechsler Intelligence Scale for Children Fourth Edition. Cross‐sectional area, volume, fractional anisotropy and radial diffusivity of the corpus callosum and five subdivisions were measured. Multivariable regression was used to assess associations between total motor score, full‐scale IQ (FSIQ) and imaging metrics. Results Adjusting for age, sex and intracranial volume, cases (N = 40) compared to controls (N = 39) demonstrated reduced whole corpus callosum area (β = −26.9, 95% confidence interval [CI] = −53.17, −0.58), volume (β = −138.5, 95% CI = −267.54, −9.56), fractional anisotropy and increased radial diffusivity (P < 0.05) within segments II–V. In cases, segment V area (β = 0.18, 95% CI = 0.004, 0.35), volume (β = 0.04, 95% CI = 0.001, 0.079), whole corpus callosum fractional anisotropy (β = 13.8 95% CI = 0.6, 27.1) and radial diffusivity (β = −11.3, 95% CI = −22.22, −0.42) were associated with FSIQ. Growth of the corpus callosum was restricted in cases with a FSIQ ≤85, and volume was reduced in cases with mild neonatal multifocal injury compared to white matter injury alone. Interpretation Following neonatal HIE, morphological and microstructural changes in the corpus callosum are associated with reduced cognitive function at early school age.

Dynamic functional connectivity (dFC) analysis of resting-state fMRI data is commonly per- formed by calculating sliding-window correlations (SWC), followed by k-means clustering in order to assign each window to a given state. Studies using synthetic data have shown that k-means per- formance is highly dependent on sliding window parameters and signal-to-noise ratio. Additionally, sources of heterogeneity between subjects may affect the accuracy of group-level clustering, thus affecting measurements of dFC state temporal properties such as dwell time and fractional occu- pancy. This may result in spurious conclusions regarding differences between groups (e.g. when comparing a clinical population to healthy controls). Therefore, is it important to quantify the ability of k-means to estimate dFC state temporal properties when applied to cohorts of multiple subjects, and to explore ways in which clustering performance can be maximised. Here, we explore the use of dimensionality reduction methods prior to clustering in order to map high-dimensional data to a lower dimensional space, providing salient features to the subse- quent clustering step. We assess the use of deep autoencoders for feature selection prior to applying k-means clustering to the encoded data. We compare this deep clustering method to feature selec- tion using principle component analysis (PCA), uniform manifold approximation and projection (UMAP), as well as applying k-means to the original feature space using either L1 or L2 distance. We provide extensive quantitative evaluation of clustering performance using synthetic datasets, representing data from multiple heterogeneous subjects. In synthetic data we find that deep clus- tering gives the best performance, while other approaches are often insufficient to capture temporal properties of dFC states. We then demonstrate the application of each method to real-world data from human subjects and show that the choice of feature selection method has a significant effect on group-level measurements of state temporal properties. We therefore advocate for the use of deep clustering as a precursor to clustering in dFC. Competing Interest Statement The authors have declared no competing interest.