MEMaría-José Escobar

María-José Escobar

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


María-José Escobar has a total of 65 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Attention maps reveal stimulus-dependent retinal population codes

    by Francisco Miqueles, Adrián G. Palacios, John Atkinson, María-José Escobar

    Introduction: Understanding how deep learning models map neural population activity to stimuli requires both high predictive accuracy and interpretable internal mechanisms. Methods: In this work, we employ the POYO framework, a scalable transformer architecture based on spike tokenization and latent modeling, to decode large-scale retinal ganglion cell recordings. We ask whether the model's attention mechanisms can provide biologically meaningful insight by evaluating two contrasting conditions: a uniform flash stimulus and a spatiotemporally structured moving ball stimulus. Results: We show that the model decodes both stimuli reliably and adapts rapidly to new preparations via fine-tuning, suggesting the capture of transferable population codes. We then analyze the model's internal organization, revealing that encoder attention patterns adapt to stimulus complexity: attention heads appear synchronized and broadly distributed for the flash stimulus, whereas they exhibit heterogeneous, specialized allocation strategies for the moving ball. By aggregating attention weights to identify the most relevant neurons for each task, we demonstrate that these high-attention units possess distinct physiological signatures—concentrating sustained, high-firing rates responses for the flash vs. diverse kinetics for the structured input. We confirm the causal validity of these findings via attention-guided ablations, where the progressive removal of these top-ranked units yields systematic losses in decoding performance. Furthermore, we expand the analysis to the decoder's attention, uncovering stimulus-specific retrieval strategies where individual heads exhibit distinct directional tuning preferences. Discussion: We conclude that generic attention mechanisms can spontaneously recover biological coding strategies, identifying functionally distinct neural subpopulations without supervision, thus validating the utility of transformer-based architectures for neuroscientific discovery.

  • Open Access

    Neural field model simulation of changes in cortical slow-wave sleep events by sensory stimulation in plasticity-evolving NREM sleep

    by Felipe A. Torres, Patricio Orio, María-José Escobar

    Brain activity during sleep transits between discretely described stages according to different brain dynamics. In particular, the slow-wave sleep stage (N2 and N3) presents events of interest for the memory consolidation process, such as slow oscillations and sleep spindles. Evidence exists that adequate timing between a slow oscillation and a sleep spindles, together with the increase in number and power, correlates with the performance in memory consolidation tasks. Nevertheless, how the results depend on the ongoing sleep stage is unclear, considering that stimulation is typically applied only at the N3 stage. Then, the effects of stimuli application in sleep macro-architecture during the same night are not well understood. Using a mean-field model of the thalamocortical activity, including dynamics of sleep cycle, we test the effects of stimuli application on the occurrence of slow-wave events. We evaluated three stimulation conditions in different time windows transiting from the N2 to N3 sleep stage. In our findings, closed-loop stimulation elicits higher quantity of slow oscillations and sleep spindles for near N3 brain dynamic. Further, closed-loop stimulation increases the co-occurrence of slow oscillations and sleep spindles. Interestingly, stimuli application does not significantly impact the trajectory of sleep dynamics, concordant with the null effect in sleep macro-architecture reported from experimental results. Finally, we noted that the lack of refractory periods in intra-thalamic dynamics results in a high occurrence of sleep spindles with stimuli application, which is not observed in the experimental results.

  • Special Issue on Emerging Topics on Development and Learning

    by Maria-Jose Escobar, Nicolas Navarro-Guerrero, Javier Ruiz-Del-Solar, Giulio Sandini

    This special issue includes state-of-the-art research on emerging topics on development and learning in natural and artificial systems. In addition to new submissions, the special issue includes extensions of the paper awarded the Best Paper Award at ICDL-EpiRob 2020—the 10th Joint IEEE Conference on Development and Learning and Epigenetic Robotics 2020 (https://cdstc.gitlab.io/icdl-2020/).


  • Bio-inspired Robotics

    by María-José Escobar, Frédéric Alexandre, Thierry Viéville, Adrian Palacios


  • Open Access

    Modelling neural entrainment and its persistence: influence of frequency of stimulation and phase at the stimulus offset

    by Mónica Otero, Caroline Lea-Carnall, Pavel Prado, María-José Escobar, Wael El-Deredy

    The entrainment (synchronization) of brain oscillations to the frequency of sensory stimuli is a key mechanism that shapes perceptual and cognitive processes, such that atypical neural entrainment leads to neuro-psychological deficits. Objective: We investigated the dynamic of neural entrainment. Particular attention was paid to the oscillatory behavior that succeed the end of the stimulation, since the persistence (reverberation) of neural entrainment may condition future sensory representations based on predictions about stimulus rhythmicity. Approach: A modified Jansen-Rit neural mass model of coupled cortical columns generated a time series whose frequency spectrum resembled that of the electroencephalogram. We evaluated spectro-temporal features of entrainment, during and after rhythmic stimulation of different frequencies, as a function of the resonance frequency of the neural population and the coupling strength between cortical columns. We tested if the duration of the entrainment persistence depended on the state of the neural network at the time the stimulus ends. Main Results: The entrainment of the column that received the stimulation was maximum when the frequency of the entrainer was within a narrow range around the resonance frequency of the column. When this occurred, entrainment persisted for several cycles after the stimulus terminated, and the propagation of the entrainment to other columns was facilitated. Propagation depended on the resonance frequency of the second column, and the coupling strength between columns. The duration of the persistence of the entrainment depended on the phase of the neural oscillation at the time the entrainer terminated, such that falling phases (from π /2 to 3 π /2 in a sine function) led to longer persistence than rising phases (from 0 to π /2 and 3 π /2 to 2 π ). Significance: The study bridges between models of neural oscillations and empirical electrophysiology, and provides insights to the use of rhythmic sensory stimulation for neuroenhancement.


  • Open Access

    Scalable and accurate method for neuronal ensemble detection in spiking neural networks

    by Rubén Herzog, Arturo Morales, Soraya Mora, Joaquín Araya, María-José Escobar, Adrian G. Palacios, Rodrigo Cofré

    We propose a novel, scalable, and accurate method for detecting neuronal ensembles from a population of spiking neurons. Our approach offers a simple yet powerful tool to study ensemble activity. It relies on clustering synchronous population activity (population vectors), allows the participation of neurons in different ensembles, has few parameters to tune and is computationally efficient. To validate the performance and generality of our method, we generated synthetic data, where we found that our method accurately detects neuronal ensembles for a wide range of simulation parameters. We found that our method outperforms current alternative methodologies. We used spike trains of retinal ganglion cells obtained from multi-electrode array recordings under a simple ON-OFF light stimulus to test our method. We found a consistent stimuli-evoked ensemble activity intermingled with spontaneously active ensembles and irregular activity. Our results suggest that the early visual system activity could be organized in distinguishable functional ensembles. We provide a Graphic User Interface, which facilitates the use of our method by the scientific community.


  • Open Access

    Selection of stimulus parameters for enhancing slow wave sleep events with a neural-field theory thalamocortical model

    by Felipe A. Torres, Patricio Orio, María-José Escobar

    Slow-wave sleep cortical brain activity, conformed by slow-oscillations and sleep spindles, plays a key role in memory consolidation. The increase of the power of the slow-wave events, obtained by auditory sensory stimulation, positively correlates with memory consolidation performance. However, little is known about the experimental protocol maximizing this effect, which could be induced by the power of slow-oscillation, the number of sleep spindles, or the timing of both events’ co-occurrence. Using a mean-field model of thalamocortical activity, we studied the effect of several stimulation protocols, varying the pulse shape, duration, amplitude, and frequency, as well as a target-phase using a closed-loop approach. We evaluated the effect of these parameters on slow-oscillations (SO) and sleep-spindles (SP), considering: (i) the power at the frequency bands of interest, (ii) the number of SO and SP, (iii) co-occurrences between SO and SP, and (iv) synchronization of SP with the up-peak of the SO. The first three targets are maximized using a decreasing ramp pulse with a pulse duration of 50 ms. Also, we observed a reduction in the number of SO when increasing the stimulus energy by rising its amplitude. To assess the target-phase parameter, we applied closed-loop stimulation at 0°, 45°, and 90° of the phase of the narrow-band filtered ongoing activity, at 0.85 Hz as central frequency. The 0° stimulation produces better results in the power and number of SO and SP than the rhythmic or random stimulation. On the other hand, stimulating at 45° or 90° change the timing distribution of spindles centers but with fewer co-occurrences than rhythmic and 0° phase. Finally, we propose the application of closed-loop stimulation at the rising zero-cross point using pulses with a decreasing ramp shape and 50 ms of duration for future experimental work.


  • Open Access

    Parallel processing of natural images by overlapping retinal neuronal ensembles

    by Jesús Pérez-Ortega, Joaquín Araya, Cristobal Ibaceta, Rubén Herzog, María-José Escobar, Fernando Peña-Ortega, Luis Carrillo-Reid, Adrian G. Palacios

    Even though the retinal microcircuit organization has been described in detail at the single-cell level, little is known about how groups of retinal cells’ coordinated activity encode and process parallel information representing the spatial and temporal structure of changing environmental conditions. To describe the population dynamics of retinal neuronal ensembles, we used microelectrode array recordings that describe hundreds of retinal ganglion cells’ simultaneous activity in response to a short movie captured in the natural environment where our subject develops their visual behaviors. The vectorization of population activity allowed the identification of retinal neuronal ensembles that synchronize to specific segments of natural stimuli. These synchronous retinal neuronal ensembles were reliably activated by the same stimuli at different trials, indicating a robust population response of retinal microcircuits. The generation of asynchronous events required integrating a physiologically meaningful time window larger than 80 ms, demonstrating that retinal neuronal ensembles’ time integration filters non-structured visual information. Interestingly, individual neurons could be part of several ensembles indicating that parallel circuits could encode environmental conditions changes. We conclude that parallel neuronal ensembles could represent the functional unit of retinal computations and propose that the further study of retinal neuronal ensembles could reveal emergent properties of retinal circuits that individual cells’ activity cannot explain.


  • Open Access

    Selection of stimulus parameters for enhancing slow wave sleep events with a Neural-field theory thalamocortical computational model

    by Felipe A. Torres, Patricio Orio, María-José Escobar

    Slow-wave sleep cortical brain activity, conformed by slow-oscillations and sleep spindles, plays a key role in memory consolidation. The increase of the power of the slow-wave events, obtained by auditory sensory stimulation, positively correlates to memory consolidation performance. However, little is known about the experimental protocol maximizing this effect, which could be induced by the power of slow-oscillation, the number of sleep spindles, or the timing of both events’ co-occurrence. Using a mean-field model of thalamocortical activity, we studied the effect of several stimulation protocols, varying the pulse shape, duration, amplitude, and frequency, as well as a target-phase using a closed-loop approach. We evaluated the effect of these parameters on slow-oscillations (SO) and sleep-spindles (SP), considering: (i) the power at the frequency bands of interest, (ii) the number of SO and SP, (iii) co-occurrences between SO and SP, and (iv) synchronization of SP with the up-peak of the SO. The first three targets are maximized using a decreasing ramp pulse with a pulse duration of 50 ms. Also, we observed a reduction in the number of SO when increasing the stimulus energy by rising its amplitude. To assess the target-phase parameter, we applied closed-loop stimulation at 0º, 45º, and 90º of the phase of the narrow-band filtered ongoing activity, at 0.85 Hz as central frequency. The 0º stimulation produces better results in the power and number of SO and SP than the rhythmic or aleatory stimulation. On the other hand, stimulating at 45º or 90º change the timing distribution of spindles centers but with fewer co-occurrences than rhythmic and 0º phase. Finally, we propose the application of closed-loop stimulation at the rising zero-cross point using pulses with a decreasing ramp shape and 50 ms of duration for future experimental work. Author summary During the non-REM (NREM) phase of sleep, events that are known as slow oscillations (SO) and spindles (SP) can be detected by EEG. These events have been associated with the consolidation of declarative memories and learning. Thus, there is an ongoing interest in promoting them during sleep by non-invasive manipulations such as sensory stimulation. In this paper, we used a computational model of brain activity that generates SO and SP, to investigate which type of sensory stimulus –shape, amplitude, duration, periodicity– would be optimal for increasing the events’ frequency and their co-occurrence. We found that a decreasing ramp of 50 ms duration is the most effective. The effectiveness increases when the stimulus pulse is delivered in a closed-loop configuration triggering the pulse at a target phase of the ongoing SO activity. A desirable secondary effect is to promote SPs at the rising phase of the SO oscillation.


  • Open Access

    Scalable and accurate automated method for neuronal ensemble detection in spiking neural networks

    by Rubén Herzog, Arturo Morales, Soraya Mora, Joaquin Araya, María-José Escobar, Adrian G. Palacios, Rodrigo Cofré

    We propose a novel, scalable, and accurate automated method for detecting neuronal ensembles from a population of spiking neurons. Our approach offers a simple yet powerful tool to study ensemble activity. It allows the participation of neurons in different ensembles, has few parameters to tune and is computationally efficient. We used spike trains of retinal ganglion cells obtained from multi-electrode array recordings under a simple ON-OFF light stimulus to test our method. We found a consistent stimuli-evoked ensemble activity intermingled with spontaneously active ensembles and irregular activity. Our results suggest that the early visual system activity is already organized in clearly distinguishable functional ensembles. To validate the performance and generality of our method, we generated synthetic data, where we found that our method accurately detects neuronal ensembles for a wide range of simulation parameters. Additionally, we found that our method outperforms current alternative methodologies. Finally, we provide a Graphic User Interface, which aims to facilitate our method’s use by the scientific community. Author summary Neuronal ensembles are strongly interconnected groups of neurons that tend to fire together (Hebb 1949). However, even when this concept was proposed more than 70 years ago, only recent advances in multi-electrode arrays and calcium imaging, statistical methods, and computing power have made it possible to record and analyze multiple neurons’ activities spiking simultaneously, providing a unique opportunity to study how groups of neurons form ensembles spontaneously and under different stimuli scenarios. Using our method, we found that retinal ganglion cells show a consistent stimuli-evoked ensemble activity, and, when validated with synthetic data, the method shows good performance by detecting the number of ensembles, the activation times, and the core-cells for a wide range of firing rates and number of ensembles accurately.