Biology subjects

Faes, L.

Publications and source records attributed to Faes, L..

14 recordsLinked to original sources

bioRxiv2026

Information-Theoretic Functional Connectivity Characterizes Multiscale Network Reorganization in Postoperative Cognitive Decline

Postoperative cognitive decline (POCD) after coronary artery bypass grafting (CABG) is increasingly conceptualized as a system-level disturbance of large-scale brain coordination rather than focal dysfunction. Here, we propose a multiscale neural engineering framework that combines static and dynamic information-theoretic connectivity with graph-theoretical analysis to characterize postoperative network vulnerability and its association with cognitive outcome. Resting-state fMRI was acquired in 14 male CABG patients at an early postoperative baseline (BL) and at 3-month follow-up (FU). Cognitive outcome at follow-up was assessed with the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS), classifying 7 patients as POCD (RBANS < 80) and 7 as NO POCD. Functional connectivity between 32 brain regions, grouped in 8 resting-state networks (RSN), was estimated using mutual information (MI; static dependence) and mutual information rate (MIR; dynamic information exchange), each computed with parametric Gaussian (linear) and model-free k-nearest neighbor estimators. Pairwise connections were validated via surrogate testing, and group differences in longitudinal connectivity change ({Delta} = FU-BL) were assessed with permutation tests at global, intra- and inter-RSN scales. Graph metrics were computed on statistically thresholded weighted networks and related to RBANS using permutation-based Spearman correlations. POCD was not associated with a uniform reduction in connectivity but with a structured pattern of network reorganization. Static connectivity showed widespread alterations, particularly within higher-order associative systems, including salience, dorsal attention, and default mode networks. Dynamic connectivity did not exhibit global group differences but revealed selective, network-specific alterations in temporal information exchange. Longitudinal analyses showed that better cognitive outcomes were associated with increased global efficiency and density and reduced modularity and small-worldness, indicating a greater brain integration. In contrast, poorer outcomes were associated with increased segregation and higher betweenness centrality, suggesting greater reliance on hub-mediated communication. Linear measures captured more widespread connectivity changes, whereas nonlinear estimators revealed more selective alterations in dynamic information flow. Combining static and dynamic information measures with complementary estimators and surrogate-validated graph analysis reveals dissociable signatures of postoperative network dysfunction. POCD is characterized by impaired restoration of distributed integration and a progressive shift toward hub-dependent communication, suggesting that large-scale integrative vulnerability may constitute a candidate biomarker of cognitive resilience after cardiac surgery.

bioengineering↗

bioRxiv2025

Direct causality measures unravel complex networks of cardiovascular oscillations and their modifications with postural stress

This study provides a comprehensive investigation of the spontaneous short-term regulatory mechanisms affecting cardiovascular and cardiorespiratory interactions during supine rest and in response to postural stress. The direct causality measure of conditional transfer entropy was applied to beat-to-beat heart period, arterial pressure, respiration, and arterial compliance variability series assessed in thirty-nine healthy subjects during the supine resting state and the orthostatic challenge. The inferred physiological networks behind these two conditions reveal well-known regulatory mechanisms, such as the tilt-induced decreased respiratory sinus arrhythmia (RSA) and increased baroreflex, as well as less explored interactions such as those involving compliance, which suggest striking physiological responses. Specifically, we found tight relationships between compliance and heart period, arterial pressure and respiration, which advocate the non negligible involvement of this cardio-vascular parameter into the intricate hank of the most studied physiological interconnections. Furthermore, the joint use of parametric and model-free estimation approaches allowed us to infer the prevalence of linear and nonlinear dynamics, as well as the effects on the inferred directed links of low- and high-frequency oscillations reflecting autonomic modulation. In conclusion, our study proves that direct causality measures are crucial to assess the characteristic links of complex cardiovascular networks and infer the many underlying short-term regulatory mechanisms. NEW & NOTEWORHTYWhile short-term regulatory mechanisms involving heart period, respiration and arterial pressure have been widely investigated, the way they produce and buffer cardiovascular oscillations in different physiological states is not fully understood. This study proposes a thorough investigation of a four-node physiological network, including the less explored arterial compliance variability, and provides insights into the linear vs. nonlinear characterization and spectral content of the causal dynamics representing each link within the network.

bioengineering↗

bioRxiv2024

Spanning spatial scales with functional imaging in the human brain; initial experiences at 10.5 Tesla

One of the most important frontiers in the efforts to improve functional imaging of brain activity (fMRI) is the recent push to increase the magnetic fields for human imaging beyond 10 Tesla. Having established safety, large gains in signal-to-noise-ratio (SNR), and novel radiofrequency arrays that capture the higher SNR, here we demonstrate major gains in the spatial resolution, sensitivity, and functional contrast of BOLD (Blood Oxygenation Level Dependent) based human fMRI at 10.5 Tesla. Using image reconstruction methods developed to minimize blurring, we also demonstrate that ultrahigh resolutions achieved at 10.5 Tesla suppress large-vein confounds in gradient-recalled-echo BOLD fMRI, thus improving the fidelity of functional signals relative to neuronal activity and yielding accurate cortical depth-profiles for layer-specific activation. Together, these multiplicative benefits deliver much needed increases in precision and resolution for meso-scale fMRI applications and illustrate the transformative potential of human functional imaging at magnetic fields that exceed 10 Tesla.

neuroscience↗

bioRxiv2024

Exploring transient neurophysiological states through local and time-varying measures of Information Dynamics

Studying the temporal evolution of complex systems requires tools able to detect the presence and quantify the strength of predictable dynamics within their output signals. Information theory aids in such a description, particularly through information storage (IS), which reflects the regularity of system dynamics by measuring the information shared between the present and the past system states. While the conventional IS computation provides an overall measure of predictable information, transient behaviors of predictability occurring during system transitions can be assessed by time-resolved measures such as the local information storage (L-IS), assuming stationarity, and the time-varying information storage (TV-IS), without stationarity assumptions. In this work, through a comparative analysis in simulated and real contexts, we aim to demonstrate how these methods complement each other and reveal dynamic changes of the system behavior associated to state transitions. In simulations, we show that the TV-IS can effectively track sudden changes of the information stored in the system, which is reflected in its average value computed over specific time intervals; on the other hand, the surprise originated by the emergence of a change in the predictability of the system is reflected in the variance of the L-IS computed within specific time intervals. In neurophysiological applications, the distinct phenomena of respiratory activity in sleep apnea and brain activity during somatosensory stimulation both reveal a significant decrease of IS evoked by state transitions, highlighting how such transitions can inject new information in physiological systems, affecting significantly their internal dynamics. Overall, TV-IS and L-IS appear to provide different and complementary information about the behavior of the systems under investigation, thereby offering valuable tools for the study of complex physiological systems where both stationary and non-stationary conditions may be present.

bioengineering↗

bioRxiv2023

Local and Global Measures of Information Storage for the Assessment of Heartbeat-Evoked Cortical Responses

ObjectiveBrain-heart interactions involve bidirectional effects produced by bottom-up input at each heartbeat, and top-down neural regulatory responses of the brain. While the cortical processing of the heartbeat is usually investigated through the analysis of the Heartbeat Evoked Potential, in this study we propose an alternative approach based on the variability in the predictability of the brain dynamics induced by the heartbeat. MethodsIn a group of eighteen subjects in whom simultaneous recording of the electroencephalogram (EEG) and electrocardiogram was performed in a resting-state, we analyzed the temporal profile of the local Information Storage (IS) to detect changes in the regularity of EEG signals in time windows associated with different phases of the cardiac cycle at rest. ResultsThe average values of the local IS were significantly higher in the parieto-occipital areas of the scalp, suggesting an activation of the Default Mode Network, regardless of the cardiac cycle phase. In contrast, the variability of the local IS showed marked differences across the cardiac cycle phases. ConclusionOur results suggest that cardiac activity influences the predictive information of EEG dynamics differently in the various phases of the cardiac cycle. SignificanceThe variability of local IS measures can represent a useful index to identify spatio-temporal dynamics within the neurocardiac system, which generally remain overlooked by the more widely employed global measures.

neuroscience↗

bioRxiv2022

Network plasticity in the honey bee antennal lobe during odor presentation

Odorant processing presents multiple parallels across animal species, and insects became relevant models for the study of olfactory coding because of the tractability of the underlying neural circuits. Within the insect brain, odorants are received by olfactory sensory neurons and processed by the antennal lobe network. Such a network comprises multiple nodes, named glomeruli, that receive sensory information and are interconnected by local interneurons participating in shaping the neural representation of an odorant. The study of functional connectivity between the nodes of a sensory network in vivo is a challenging task that requires simultaneous recording from multiple nodes at high temporal resolutions. Here, we used fast two-photon microscopy to follow the calcium dynamics of antennal lobe glomeruli and applied Granger causality analysis to assess the functional directed connectivity among network nodes during and after olfactory stimulation. Our findings show that there is a significant causal coupling between network nodes also in the absence of olfactory stimulation. However, connectivity patterns change upon odorant arrival, showing an increase in connection density and conveying odorant-specific information. Furthermore, we show that causal functional connectivity is unique in individual subjects, and cannot be detected upon artificial glomerular scrambling across individuals. This demonstrates that network states do not consist solely of correlations of mean glomerular responses, which are known to be conserved across bees.

neuroscience↗

bioRxiv2022

Feasibility of Ultra-Short Term Analysis of Heart Rate and Systolic Arterial Pressure Variability at Rest and During Stress via Time-domain and Entropy-based Measures

Heart Rate Variability (HRV) and Blood Pressure Variability (BPV) are widely employed tools for characterizing the complex behavior of cardiovascular dynamics. Usually, HRV and BPV analyses are carried out through short-term (ST) measurements, which exploit [~]5 minute-long recordings. Recent research efforts are focused on reducing the time series length, assessing whether and to what extent Ultra-Short Term (UST) analysis is capable of extracting information about cardiovascular variability from very short recordings. In this work, we compare ST and UST measures computed on electrocardiographic R-R intervals and systolic arterial pressure time series obtained at rest and during both postural and mental stress. Standard time-domain indices are computed, together with entropy-based measures able to assess regularity and complexity of cardiovascular dynamics, on time series lasting up to 60 samples, employing either a faster linear parametric estimator or a more reliable but time-consuming model-free method based on nearest neighbor estimates. Our results evidence that shorter time series up to 120 samples still exhibit an acceptable agreement with the ST reference, and can be exploited to discriminate between stress and rest as well. Moreover, although neglecting nonlinearities inherent to short-term cardiovascular dynamics, the faster linear estimator is still capable of detecting differences among the conditions, thus resulting suitable to be implemented on wearable devices.

bioengineering↗

bioRxiv2020

A measure of concurrent neural firing activity based on mutual information

Multiple methods have been developed in an attempt to quantify stimulus-induced neural coordination and to understand internal coordination of neuronal responses by examining the synchronization phenomena in neural discharge patterns. In this work we propose a novel approach to estimate the degree of concomitant firing between two neural units, based on a modified form of mutual information (MI) applied to a two-state representation of the firing activity. The binary profile of each single unit unfolds its discharge activity in time by decomposition into the state of neural quiescence/low activity and state of moderate firing/bursting. Then, the MI computed between the two binary streams is normalized by their minimum entropy and is taken as positive or negative depending on the prevalence of identical or opposite concomitant states. The resulting measure, denoted as Concurrent Firing Index based on MI (CFIMI), relies on a single input parameter and is otherwise assumption-free and symmetric. Exhaustive validation was carried out through controlled experiments in three simulation scenarios, showing that CFIMI is independent on firing rate and recording duration, and is sensitive to correlated and anti-correlated firing patterns. Its ability to detect non-correlated activity was assessed using ad-hoc surrogate data. Moreover, the evaluation of CFIMI on experimental recordings of spiking activity in retinal ganglion cells brought insights into the changes of neural synchrony over time. The proposed measure offers a novel perspective on the estimation of neural synchrony, providing information on the co-occurrence of firing states in the two analyzed trains over longer temporal scales compared to existing measures.

neuroscience↗

bioRxiv2020

Information Decomposition in the Frequency Domain: a New Framework to Study Cardiovascular and Cardiorespiratory Oscillations

While cross-spectral and information-theoretic approaches are widely used for the multivariate analysis of physiological time series, their combined utilization is far less developed in the literature. This study introduces a framework for the spectral decomposition of multivariate information measures, which provides frequency-specific quantifications of the information shared between a target and two source time series and of its expansion into amounts related to how the sources contribute to the target dynamics with unique, redundant and synergistic information. The framework is illustrated in simulations of linearly interacting stochastic processes, showing how it allows to retrieve amounts of information shared by the processes within specific frequency bands which are otherwise not detectable by time-domain information measures, as well as coupling features which are not detectable by spectral measures. Then, it is applied to the time series of heart period, systolic and diastolic arterial pressure and respiration variability measured in healthy subjects monitored in the resting supine position and during head-up tilt. We show that the spectral measures of unique, redundant and synergistic information shared by these variability series, integrated within specific frequency bands of physiological interest, reflect the mechanisms of short term regulation of cardiovascular and cardiorespiratory oscillations and their alterations induced by the postural stress.

bioengineering↗

bioRxiv2019

Feasibility of Automated Deep Learning Design for Medical Image Classification by Healthcare Professionals with Limited Coding Experience

Deep learning has huge potential to transform healthcare. However, significant expertise is required to train such models and this is a significant blocker for their translation into clinical practice. In this study, we therefore sought to evaluate the use of automated deep learning software to develop medical image diagnostic classifiers by healthcare professionals with limited coding - and no deep learning - expertise.\n\nWe used five publicly available open-source datasets: (i) retinal fundus images (MESSIDOR); (ii) optical coherence tomography (OCT) images (Guangzhou Medical University/Shiley Eye Institute, Version 3); (iii) images of skin lesions (Human against Machine (HAM)10000) and (iv) both paediatric and adult chest X-ray (CXR) images (Guangzhou Medical University/Shiley Eye Institute, Version 3 and the National Institute of Health (NIH)14 dataset respectively) to separately feed into a neural architecture search framework that automatically developed a deep learning architecture to classify common diseases. Sensitivity (recall), specificity and positive predictive value (precision) were used to evaluate the diagnostic properties of the models. The discriminative performance was assessed using the area under the precision recall curve (AUPRC). In the case of the deep learning model developed on a subset of the HAM10000 dataset, we performed external validation using the Edinburgh Dermofit Library dataset.\n\nDiagnostic properties and discriminative performance from internal validations were high in the binary classification tasks (range: sensitivity of 73.3-97.0%, specificity of 67-100% and AUPRC of 0.87-1). In the multiple classification tasks, the diagnostic properties ranged from 38-100% for sensitivity and 67-100% for specificity. The discriminative performance in terms of AUPRC ranged from 0.57 to 1 in the five automated deep learning models. In an external validation using the Edinburgh Dermofit Library dataset, the automated deep learning model showed an AUPRC of 0.47, with a sensitivity of 49% and a positive predictive value of 52%. The quality of the open-access datasets used in this study (including the lack of information about patient flow and demographics) and the absence of measurement for precision, such as confidence intervals, constituted the major limitation of this study.\n\nAll models, except for the automated deep learning model trained on the multi-label classification task of the NIH CXR14 dataset, showed comparable discriminative performance and diagnostic properties to state-of-the-art performing deep learning algorithms. The performance in the external validation study was low. The availability of automated deep learning may become a cornerstone for the democratization of sophisticated algorithmic modelling in healthcare as it allows the derivation of classification models without requiring a deep understanding of the mathematical, statistical and programming principles. Future studies should compare several application programming interfaces on thoroughly curated datasets.

bioinformatics↗

bioRxiv2019

The Moorfields AMD Database Report 2 - Fellow Eye Involvement with Neovascular Age-related Macular Degeneration

Background/AimsNeovascular age-related macular degeneration (nAMD) is frequently bilateral, and previous reports on fellow eyes have assumed sequential treatment after a period of treatment of the first eye only. The aim of our study was to analyse baseline characteristics and visual acuity (VA) outcomes of fellow eye involvement with nAMD, specifically differentiating between sequential and non-sequential (due to macular scarring in the first eye) anti-vascular endothelial growth factor treatment and timelines for fellow eye involvement.\n\nMethodsRetrospective, electronic medical record database study of the Moorfields AMD database of 8174 eyes/120,756 single entries with data extracted between October 21, 2008 and August 9, 2018. The dataset for analysis consisted of 1180 sequential, 413 nonsequential, and 1110 unilateral eyes.\n\nResultsMean VA of sequentially treated fellow eyes at baseline was significantly higher (62{+/-}13), VA gain over two years lower (0.65{+/-}14), and proportion of eyes with good VA ([≥]20/40 or 70 letters) higher (46%) than the respective first eyes (baseline VA 54{+/-}16, VA gain at two years 5.6{+/-}15, percentage of eyes with good VA 38%). Non-sequential fellow eyes showed baseline characteristics and VA outcomes similar to first eyes. Fellow eye involvement rate was 32% at two years, and median time interval to fellow eye involvement was 71 (IQR 27-147) weeks.\n\nConclusionThis reports shows sequentially treated nAMD fellow eyes have better baseline and final VA than non-sequentially treated eyes after 2 years of treatment. Sequentially treated eyes also had a greater proportion with good VA after 2 years of treatment.\n\nPRECISDepending on age, fellow eye involvement occurs in 32% of patients with neovascular AMD by two years. Fellow eyes generally maintain better vision, except in cases where late-stage disease in the first eye was untreated.

epidemiology↗

bioRxiv2019

Information decomposition of multichannel EMG to map functional interactions in the distributed motor system

The central nervous system needs to coordinate multiple muscles during postural control. Functional coordination is established through the neural circuitry that interconnects different muscles. Here we used multivariate information decomposition of multichannel EMG acquired from 14 healthy participants during postural tasks to investigate the neural interactions between muscles. A set of information measures were estimated from an instantaneous linear regression model and a time-lagged VAR model fitted to the EMG envelopes of 36 muscles. We used network analysis to quantify the structure of functional interactions between muscles and compared them across experimental conditions. Conditional mutual information and transfer entropy revealed sparse networks dominated by local connections between muscles. We observed significant changes in muscle networks across postural tasks localized to the muscles involved in performing those tasks. Information decomposition revealed distinct patterns in task-related changes: unimanual and bimanual pointing were associated with reduced transfer to the pectoralis major muscles, but an increase in total information compared to no pointing, while postural instability resulted in increased information, information transfer and information storage in the abductor longus muscles compared to normal stability. These findings show robust patterns of directed interactions between muscles that are task-dependent and can be assessed from surface EMG recorded during static postural tasks. We discuss directed muscle networks in terms of the neural circuitry involved in generating muscle activity and suggest that task-related effects may reflect gain modulations of spinal reflex pathways.

neuroscience↗

bioRxiv2018

One and Two Year Visual Outcomes from the Moorfields AMD Database - an Open Science Resource for the Study of Neovascular Age-related Macular Degeneration

ObjectivesTo analyse treatment outcomes and share clinical data from a large, single-center, well-curated database (8174 eyes / 6664 patients with 120,756 single entries) of patients with neovascular age related macular degeneration (AMD) treated with anti-vascular endothelial growth factor (VEGF). By making our depersonalised raw data openly available, we aim to stimulate further research in AMD, as well as setting a precedent for future work in this area.\n\nSettingRetrospective, comparative, non-randomised electronic medical record (EMR) database cohort study of the UK Moorfields AMD database with data extracted between 2008 and 2018.\n\nParticipants3357 eyes/patients (61% female). Extraction criteria were [≥] 1 ranibizumab or aflibercept injection, entry of \"AMD\" in the diagnosis field of the EMR, and a minimum of one year of follow-up. Exclusion criteria were unknown date of first injection and treatment outside of routine clinical care at Moorfields before the first recorded injection in the database.\n\nMain outcome measuresPrimary outcome measure was change in VA at one and two years from baseline as measured in Early Treatment Diabetic Retinopathy Study (ETDRS) letters. Secondary outcomes were the number of injections and predictive factors for VA gain.\n\nResultsMean VA gain at one-year and two years were +5.5{+/-}0.5 and +4.9{+/-}0.68 letters respectively. Fifty-four percent of eyes gained [≥]5 letters at two years, 63% had stable VA ({+/-}[≤]14 letters), forty-four percent of eyes maintained good VA ([≥]70 letters). Patients received a mean of 7.7{+/-}0.06 injections during year one and 13.0{+/-}0.2 injections over two years.\n\nYounger age, lower baseline VA, and more injections were associated with higher VA gain at two years.\n\nConclusionThis study benchmarks high quality EMR study results of real life AMD treatment and promotes open science in clinical AMD research by making the underlying data publicly available.\n\nStrengths and limitations of this study- Large sample size, retrospective, single centre, electronic medical record database study\n- High quality real life data\n- Open science approach with sharing of depersonalised raw data

epidemiology↗

bioRxiv2018

Task-induced deactivation in diverse brain systems correlates with interindividual differences in distinct autonomic indices

Neuroimaging research has shown that different cognitive tasks induce relatively specific activation patterns, as well as less task-specific deactivation patterns. Here we examined whether individual differences in Autonomic Nervous System (ANS) activity during task performance correlate with the magnitude of task-induced deactivation. In an fMRI study, participants performed a continuous mental arithmetic task in a task/rest block design, while undergoing combined fMRI and heart / respiration rate acquisitions using photoplethysmograph and respiration belt. As expected, task performance increased heart-rate and reduced the RMSSD, a cardiac index related to vagal tone. Across participants, higher heart rate during task was linked to increased activation in fronto-parietal regions, as well as to stronger deactivation in ventromedial prefrontal regions. Respiration frequency during task was associated with similar patterns, but in different regions than those identified for heart-rate. Finally, in a large set of regions, almost exclusively limited to the Default Mode Network, lower RMSSD was associated with greater deactivation, and furthermore, the vast majority of these regions were task-deactivated at the group level. Together, our findings show that inter-individual differences in ANS activity are strongly linked to task-induced deactivation. Importantly, our findings suggest that deactivation is a multifaceted construct potentially linked to ANS control, because distinct ANS measures correlate with deactivation in different regions. We discuss the implications for current theories of cortical control of the ANS and for accounts of deactivation, with particular reference to studies documenting a \"failure to deactivate\" in multiple clinical states.

neuroscience↗