Co-Activation Patterns in Neonates using High-Density Diffuse Optical Tomography: Insights into Early Dynamic Functional Connectivity

Lee, K. E.; Austin, T.; Uchitel, J.; Caballero Gaudes, C.; Collins-Jones, L. H.; Cooper, R.; Edwards, A.; Hebden, J.; Kromm, G.; Pammenter, K.; Blanco, B.

doi:10.1101/2025.03.15.643385

bioRxiv · 10.1101/2025.03.15.643385

Co-Activation Patterns in Neonates using High-Density Diffuse Optical Tomography: Insights into Early Dynamic Functional Connectivity

Lee, K. E.·Austin, T.·Uchitel, J.·Caballero Gaudes, C.·Collins-Jones, L. H.·Cooper, R.·Edwards, A.·Hebden, J.·Kromm, G.·Pammenter, K.·Blanco, B.

View original source ↗Export metadata ↓

Abstract

SignificanceDynamic functional connectivity (FC) in neonates is a growing area of interest due to the developmental significance of early functional networks. There are several emerging techniques to measure dynamic FC, adding new perspectives to well-studied static FC networks. Recent dynamic FC studies suggest that adult resting-state networks are driven by key moments of dynamic activity rather than sustained correlations. Co-activation pattern (CAP) analysis leverages this theory, clustering high-activity frames to identify recurring configurations of significant activity. High-density diffuse optical tomography (HD-DOT) is an infant-friendly modality that measures hemodynamic changes in the cortex and has been used to investigate static FC in term-aged infants. CAP analysis has not yet been applied to neonatal HD-DOT and may reveal temporal features of early functional networks that are not apparent from static methods. AimThis study applies CAP analysis to neonatal HD-DOT to characterize transient co-activation states and provide new insight into early functional brain networks beyond what is known from static FC methods alone. ApproachTask-free HD-DOT data were acquired from a cohort of sleeping term newborns at the Rosie Hospital, Cambridge UK (n = 44, postmenstrual age = 40+3 (range: 38+2-42+6) weeks). In each recording, the top 15% of seed-selected frames were clustered using the K-means algorithm for three regions of interest (ROIs: frontal, central, and parietal) to identify significant seed-associated patterns of co-activation or co-deactivation. These co-activation patterns (CAPs) were characterized for each infant using four metrics: consistency, fractional occupancy, dwell time, and transition likelihood. ResultsDistinct CAPs, reflecting the dynamic organization of neonatal cortical networks, were identified for frontal, central, and parietal regions. These CAPs showed high consistency scores, reflecting high intra-cluster spatial correlation and validating the efficacy of CAP analysis for newborn HD-DOT data. The CAP decomposition revealed significant patterns not observed in conventional static functional connectivity analyses. Several neonatal CAPs exhibited frontoparietal co-activation, potentially reflecting early default mode network activity, which is immature and modular for the first year after birth. ConclusionsThis work demonstrates the utility of CAP analysis with newborn HD-DOT and provides new insight into the dynamics of neonatal functional connectivity.

Explore related subjects

neuroscience Preprint

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Lee, K. E., Austin, T., Uchitel, J., Caballero Gaudes, C., Collins-Jones, L. H., Cooper, R., Edwards, A., Hebden, J., Kromm, G., Pammenter, K., Blanco, B.. 2025-03-15. Co-Activation Patterns in Neonates using High-Density Diffuse Optical Tomography: Insights into Early Dynamic Functional Connectivity. https://doi.org/10.1101/2025.03.15.643385

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

CoReg, tracking-based motion compensation for in vivo calcium imaging

In vivo calcium imaging enables the observation of neural activity with high spatial and temporal resolution. However, recordings are frequently affected by motion artifacts caused by animal movement, physiological processes, tissue drift, or instability of the imaging setup. These artifacts make signal extraction harder and reduce the consistency of downstream analyses. Current registration methods can correct rigid and non-rigid motion, but often estimate motion from individual frames or image patches and make limited use of temporal information. In addition, activity-driven fluorescence changes can make it difficult to distinguish changes in image appearance from actual motion. Here, we treat motion correction in calcium imaging as a point tracking problem and introduce CoReg, an open-source Python- based software tool with a graphical user interface. We benchmark CoReg against established registration methods using fluorescence imaging datasets from multiple species. CoReg performs competitively with or better than existing approaches across several settings. We further show that its performance is robust to the choice of tracked points, making the method straightforward to apply. CoReg includes implementations of both our method and the evaluated established methods. Our results show that tracking-based registration provides a simple and reliable approach for motion correction in vivo calcium imaging.

neuroscience↗

bioRxiv2026

A shared slowing axis and phenotype-specific rhythms link clinical EEG to FDG-PET dementia phenotypes

Early and accurate identification of neurodegenerative disorders has notable implications for treatment and quality of life. Whereas neuroimaging-based biomarkers can accurately identify neurodegenerative subtypes, EEG represents a low-cost alternative that could expand access to dementia care. Here, we investigated whether six data-driven EEG latent factors, derived by nonnegative tensor decomposition of clinical EEGs, differ systematically across FDG-PET-defined neurodegenerative phenotypes. In 455 patients with EEG and FDG-PET acquired within one year of each other, factor expression was compared across 14 diagnostic classes assigned by StateViewer, a validated artificial intelligence framework for FDG-PET interpretation. To separate disease-general severity from phenotype-specific topography, each factor was also regressed on a subject-level slowing index. Finally, we related the factors to regional FDG uptake and to neuropsychological performance (n = 98) and assessed their potential utility for single-subject classification. Four factors reflecting neurophysiological oscillations (posterior alpha, Rolandic beta, frontotemporal delta-theta and posterior theta-slow alpha) varied significantly across phenotypes, with attenuation of faster and augmentation of slower rhythms most pronounced in dementia with Lewy bodies (DLB), dysexecutive Alzheimer disease (AD) and posterior cortical atrophy, whereas factors dominated by eye-movement and muscle artifact showed no robust phenotypic association. Alpha and beta effects were largely attributable to overall slowing, whereas posterior theta-slow alpha became more phenotype-specific after adjustment (eta-squared 0.082 to 0.147), dissociating the Alzheimer disease spectrum from frontotemporal syndromes. The slowing index tracked with an FDG-PET pattern highly associated with DLB across the cohort, irrespective of assigned phenotype (r = 0.51, P < 0.001). Consistent with a model in which dementia phenotypes reflect altered hippocampal-neocortical oscillatory coupling (stochastic latent oscillatory diffusion), frontotemporal delta-theta tracked a limbocortical axis of metabolic pattern expression, rising with neocortical and falling with limbic-predominant involvement, independent of overall slowing and phenotype. Each factor mapped onto a distinct metabolic topography, with frontotemporal delta-theta resembling Alzheimer-type hypometabolism and posterior theta-slow alpha the medial occipital hypometabolism of DLB. In terms of neuropsychological performance, Rolandic beta was associated with executive functions, and frontotemporal delta-theta with memory, visuospatial and global performance. Using these four factors alone, an exploratory logistic regression classifier separated cognitively unimpaired participants from DLB, posterior cortical atrophy, typical AD, and dysexecutive AD (area under the curve > 0.8) but poorly discriminated frontotemporal dementia phenotypes canonically associated with 4-repeat tau. Thus, data-driven decomposition of clinical EEG separates oscillatory activity from artifact into factors that track metabolic phenotype, regional metabolism and cognition, capturing both overall slowing and phenotype-specific patterns. This work informs the further development of scalable, biology-informed EEG-based phenotyping of neurodegenerative disease.

neuroscience↗

bioRxiv2026

An application of fNIRS for the prediction of hearing thresholds in sleeping infants: a proof of principle

This study applied a two-component-response modelling approach to fNIRS recordings obtained from naturally sleeping infants presented with auditory stimuli at different sound intensities. We investigated whether the relationship between fNIRS response measures and stimulus intensity could be modelled to estimate individual auditory hearing thresholds. In the normal-hearing group, 75% of infants had threshold estimates for the sound /ba/ within the expected normal-hearing range, with an estimated mean equivalent normal loudness (ENL) of 10.72 dB HL (SD 8.64 dB HL). A case study further explored the potential application of this approach in an infant with hearing impairment, highlighting the importance of reliable responses at lower stimulus intensities and appropriate regression model selection. These findings provide preliminary evidence that cortical haemodynamic responses can be quantitatively related to sound intensity for estimating auditory sensitivity.

neuroscience↗