Biology subjects

Guevara, P.

Publications and source records attributed to Guevara, P..

8 recordsLinked to original sources

bioRxiv2026

A supervised digital game intervention supports language and communication in young children.

The role of digital media in early childhood development remains highly debated, particularly regarding its impact on language acquisition. While excessive or unsupervised screen exposure has been linked to poorer outcomes, less is known about whether structured and interactive uses of technology can support learning. Building on previous research, we evaluated a brief, educator-supervised tablet-based intervention in 246 children aged 2-5 years from low- to middle-socioeconomic backgrounds attending public early education centers. Using a pre-post design with matched study and control groups, children completed 4-8 short training sessions (15 minutes each) involving interactive word-image associations spanning multiple linguistic categories. Preschoolers additionally engaged in prompted vocalization. Across age groups (2-3, 3-4, and 4-5 years), children in the intervention showed greater gains in language comprehension than controls, including receptive language in toddlers ({beta} = 0.49, p = 0.009), vocabulary and morphology in younger preschoolers ({beta} = 0.59-0.68, all p < 0.05), and grammar comprehension in older preschoolers ({beta} = 0.30, p = 0.038). These effects were consistent after accounting for child and parental characteristics. Together, these findings suggest that the developmental impact of digital media depends less on exposure itself than on how it is used. When embedded in structured, socially guided interactions, even brief tablet-based activities may support early language development

developmental biology↗

bioRxiv2026

NucVerse3D: Generalizable 3D nuclear instance segmentation across heterogeneous microscopy modalities

Accurate three-dimensional (3D) nuclear instance segmentation is a prerequisite for quantitative phenotyping in volumetric microscopy, yet remains challenging in densely packed tissues, irregular nuclear morphologies, and across heterogeneous imaging modalities. Here we present NucVerse3D, a deep-learning framework for generalized 3D nuclei instance segmentation that combines a residual attention 3D U-Net architecture with a reversible gradient-field representation for robust centroid-aware instance reconstruction. NucVerse3D is trained end to end in 3D using modality-agnostic preprocessing and isotropic scale normalization, enabling deployment across confocal microscopy, two-photon microscopy, light-sheet microscopy, micro-computed tomography, and scanning electron microscopy volumes. We benchmarked NucVerse3D on seven volumetric datasets spanning multiple species and tissues, comprising more than forty thousand manually annotated nuclei, including newly released ground-truth datasets of mouse liver tissue (control and hepatocellular carcinoma) and Drosophila brain glial nuclei. Across datasets, NucVerse3D achieved consistently high precision, recall, F1-score, and average precision, and outperformed the state-of-the-art methods particularly in dense and irregular settings, while remaining competitive on simpler cases. A single generalized model trained on pooled data matched the performance of dataset-specific models, and ablation experiments demonstrated that preprocessing and scale normalization substantially contribute to performance under strict intersection-over-union criteria. To demonstrate the biomedical utility of NucVerse3D, we applied it to three-dimensional liver images from a mouse model of hepatocellular carcinoma (HCC). High-fidelity, nucleus-by-nucleus segmentation enabled the quantification of the Nuclear Decoupling Score (NDS), which captures deviations in nuclear DNA-volume coupling at the single-nucleus level. NDS analysis revealed a progressive increase in nuclear abnormalities within tumor regions, forming spatially coherent domains of dysregulated nuclei and highlighting NDS as a potential quantitative biomarker of dysplastic and tumor tissue. Together, NucVerse3D provides a robust and generalizable solution for 3D nuclear instance segmentation and enables quantitative nuclear phenotyping across imaging modalities. Highlights- NucVerse3D provides accurate 3D nuclear instance segmentation across modalities - Residual attention and gradient fields enable robust separation of dense nuclei - New 3D annotated datasets of mouse liver and Drosophila brain are released - A generalized model achieves performance comparable to dataset-specific training - 3D nuclear phenotyping reveals spatially organized nuclear abnormalities in HCC

cell biology↗

bioRxiv2025

Deep White-Matter Pathways Mediate the Link Between Docosahexaenoic Acid (DHA) Status and Cognitive Performance in Adolescence

Docosahexaenoic acid (DHA) is a polyunsaturated fatty acid enriched in neuronal membranes and myelin and associated with cognitive performance. However, nutritional interventions show inconsistent cognitive effects, partly due to limited knowledge of the neural pathways linking DHA status to human cognition during sensitive periods of white-matter maturation, such as adolescence. We addressed this gap by studying 99 adolescents drawn from both extremes of performance on a national scholastic examination. Participants completed assessments of scholastic achievement (SA) and intellectual ability (IA), provided erythrocyte DHA samples, and underwent multimodal MRI, including diffusion, T1-weighted, and T2-weighted imaging. Independent component analysis and Bayesian multivariate LASSO models identified brain components jointly associated with DHA and cognition. Across four MRI modalities, a single deep white-matter component consistently emerged as the strongest shared pathway linking DHA with cognition. Tract-resolved analyses highlighted predominant contributions from the fornix and thalamus-temporal fasciculus, with additional subcortical and cortical involvement. In joint models, these components predicted SA and IA after accounting for DHA and other fatty acids, consistent with an indirect, mediation-like pathway. These findings move beyond DHA-behavior correlations by identifying specific neuroanatomical pathways through which a modifiable dietary factor relates to adolescent learning and intellectual performance, offering mechanistic insight relevant to neuroscience, nutrition, and education. Significance StatementAdolescence is a sensitive period for the maturation of white-matter pathways that support learning and reasoning. Docosahexaenoic acid (DHA), an essential dietary fatty acid enriched in neuronal membranes and myelin, has been linked to cognitive performance, yet the neural mechanisms underlying this association remain unclear. Using multimodal MRI and Bayesian multivariate modeling in adolescents with high or low scholastic performance, identify a specific deep white-matter pathway--centered on the fornix and the thalamus-temporal fasciculus--as the principal route connecting DHA status with scholastic achievement and intellectual ability. Additional subcortical and cortical contributions reveal a coordinated system-level architecture. These findings move beyond correlations by providing mechanistic insight into how a modifiable nutritional factor relates to cognitive development, with implications for neuroscience, public health, and education.

neuroscience↗

bioRxiv2025

Frequency-Dependent Neurostimulation in Whole Brain Network models

BackgroundBrain stimulation at specific frequencies has been shown to have therapeutic effects for certain neurological disorders. However, it remains unclear how these oscillatory inputs interact with neuronal dynamics and what drives this frequency selectivity. Here, we propose that this is achieved by targeting specific brain circuits through frequency-selective network entrainment. MethodsTo demonstrate that this type of entrainment can occur in the brain connectome structure, we use a minimal physics-based model of coupled oscillators that preserves only the essential structural connectivity of real brain networks. Using this model, we test how periodic stimulation influences network oscillatory dynamics, identifying sub-networks that synchronize with specific stimulation frequencies. Furthermore, we validate the suitability of the model to reproduce well-known frequency-selective entrainment using visual and auditory periodic stimulation. ResultsOur model reproduces selective network entrainment, whereby distinct stimulation frequencies favor the synchrony of different sub-networks. Crucially, entrainment exhibits a physics-predicted inverse relationship with frequency: lower-frequency stimulation favors broader synchrony across the network, while higher-frequency inputs produce spatially confined effects. These patterns emerge purely from network structure and oscillatory physics, independent of biological details. ConclusionThis physics-based approach elucidates the fundamental mechanistic principles governing frequency-selective brain network stimulation. Our minimal model demonstrates that selective entrainment of specific subnetworks can be achieved through frequency selection alone, largely determined by network structure and oscillatory dynamics. These findings provide a theoretical foundation for understanding how network architecture determines stimulation selectivity, supporting the development of principled approaches to targeted neuromodulation. Author summaryIn this work, we explore how external rhythmic stimulation interacts with the brains own rhythms. We use a physics-based model that represents the brain as a network of interconnected oscillators, linked according to the human connectome. Through this model, we show that stimulation effects are not limited to the targeted region but spread across the network, depending on the stimulation frequency, its spatial location and the underlying network dynamics. Our results reveal that low-frequency rhythms promote large-scale synchrony, while high-frequency rhythms engage more localized areas, all emerging from the brains structural connections and communication delays. The simulated cortical activation exhibits a frequency-selective profile that aligns strongly with that estimated from sensory-evoked electroencephalographic activity. This agreement suggests that differences between individuals in how their brain networks operate may explain why stimulation affects each person differently. Our findings offer a mechanistic framework for designing personalized stimulation protocols, illustrating how computational modeling can guide the optimization of noninvasive brain stimulation in both research and clinical contexts.

bioengineering↗

bioRxiv2024

Neural correlates in the time course of inferences: costs and benefits for less-skilled readers at university level.

Inferences are an indicator of a greater reading comprehension, as they imply a combination of implicit and explicit information that usually combines a textual representation with background knowledge of the reader. The aim of this study is to explore the costs and benefits of the time course of inferences in university students with reading comprehension difficulties at 3 stages during a narration. The method used was the event-related potential (ERP) technique in order to register the brain activity of 63 teaching program students while they read familiar, less-familiar and neutral stories. Results show a slow negativity potential component with greater negativity in words coming from familiar contexts when compared to less familiar and neutral ones in the first locus; an N400 component and a Post-N400 component in the second locus, reflecting greater negativity in familiar contexts when compared to less-familiar ones; and, lastly, through the use of a lexical decision task, FN400 and N400 components were found in the third locus, especially for pseudowords. These results are interpreted as a preferably bottom-up processing, which is characterized by lexical access difficulties in less-skilled readers.

neuroscience↗

bioRxiv2023

Virtual tissue microstructure reconstruction by deep learning and fluorescence microscopy

Analyzing tissue microstructure is essential for understanding complex biological systems in different species. Tissue functions largely depend on their intrinsic tissue architecture. Therefore, studying the three-dimensional (3D) microstructure of tissues, such as the liver, is particularly fascinating due to its conserved essential roles in metabolic processes and detoxification. Here, we present TiMiGNet, a novel deep learning approach for virtual 3D tissue microstructure reconstruction using Generative Adversarial Networks (GANs) and fluorescence microscopy. TiMiGNet overcomes challenges such as poor antibody penetration and time-intensive procedures by generating accurate, high-resolution predictions of tissue components across large volumes without the need of paired images as input. We applied TiMiGNet to analyze tissue microstructure in mouse and human liver tissue. TiMiGNet shows high performance in predicting structures like bile canaliculi, sinusoids, and Kupffer cell shapes from actin meshwork images. Remarkably, using TiMiGNet we were able to computationally reconstruct tissue structures that cannot be directly imaged due experimental limitations in deep dense tissues, a significant advancement in deep tissue imaging. Our open-source virtual prediction tool facilitates accessible and efficient multi-species tissue microstructure analysis, accommodating researchers with varying expertise levels. TiMiGNets simplicity and independence from paired images make it a versatile asset. Overall, our method represents a powerful and efficient approach for studying tissue microstructure, with far-reaching applications in diverse biological contexts and species.

bioinformatics↗

bioRxiv2022

Efficient estimation of time-dependent functional connectivity using Structural Connectivity constraints

Multivariate autoregressive models [MAR] allows estimating effective brain connectivity by considering both power and phase fluctuations of the signals involved. A MAR models brain activity in one region as a linear combination of past activations in all other regions. A Hidden Markov model, HMM, whose states emisions are drawn from state-specific MARs, can then be used to model fast switching of effective brain connectivity over time. However, the large number of MAR parameters, impede the accurate and efficient estimation of such models from neuroimaging timeseries with limited length. We propose a new model for inferring time-dependent effective brain connectivity by using a sparse MAR parameterisation to model the states emisions of a Hidden Semi-Markov Model, HsMM-MAR-AC. The sparse MAR model parameters are restricted by Anatomical Connectivity information in two ways: direct effective connectivity between two regions is only considered if the corresponding structural link/connection exists; and the time-lag associated with each direct connection is computed based on the average fibre length between the two regions, such that only one lag per connection is estimated. We simulated ground truth time-dependent brain connectivity states by generating time-series of 4 to 10 minutes sampled at 5ms, from switching Resting State Networks with a reference structural connectivity, and evaluated the accuracy of the new model in recovering the simulated Brain connectivity States against different levels of connectivity thresholding above and below the reference. We show that even when restricting the MAR to half of the reference connections, the model was able to recover the number of brain states, the associated connectivity features and their dynamics, with as little data as 4 mins. More relaxed structural connectivity thresholds required longer data to estimate the model accurately, and became computationally unfeasible without anatomical restrictions. HsMM-MAR-AC offers an efficient algorithm for estimating time-depend Effective Connectivity (tdEC) from neuroimaging data, that exploits the advantages of MAR without identifiability problems, excessive demand on data collection, or unnecessary computational complexity.

bioinformatics↗

bioRxiv2018

Increased and decreased superficial white matter structural connectivity in schizophrenia and bipolar disorder

Schizophrenia (SZ) and bipolar disorder (BD) are often conceptualized as \"disconnection syndromes\", with substantial evidence of abnormalities in deep white matter tracts, forming the substrates of long-range connectivity, seen in both disorders. However, the study of superficial white matter (SWM) U-shaped short-range tracts remained challenging until recently, although findings from post-mortem studies suggest they are likely integral components of SZ and BD neuropathology. This diffusion weighted imaging (DWI) study aimed to investigate SWM microstructure in vivo in both SZ and BD for the first time. We performed whole brain tractography in 31 people with SZ, 32 people with BD and 54 controls using BrainVISA and Connectomist 2.0. Segmentation and labelling of SWM tracts were performed using a novel, comprehensive U-fiber atlas. Analysis of covariances yielded significant generalized fractional anisotropy (gFA) differences for 17 SWM bundles in frontal, parietal and temporal cortices. Post hoc analyses showed gFA reductions in both patient groups as compared with controls in bundles connecting regions involved in language processing, mood regulation, working memory and motor function (pars opercularis, insula, anterior cingulate, precentral gyrus). We also found increased gFA in SZ patients in areas overlapping the default mode network (inferior parietal, middle temporal, precuneus), supporting functional hyperconnectivity of this network evidenced in SZ. We thus illustrate that short U-fibers are vulnerable to the pathological processes in major psychiatric illnesses, encouraging improved understanding of their anatomy and function.

neuroscience↗