Biology subjects

Lebedev, M. A.

Publications and source records attributed to Lebedev, M. A..

13 recordsLinked to original sources

bioRxiv2026

Cross-Recording Handwritten Digit Decoding from sEMG Using a Compact CNN-Transformer and Few-Shot Adaptation

Surface electromyography (sEMG) offers a silent and wearable input modality, but its practical use is limited by variability across users and recording sessions. This study presents a compact CNN- Transformer model for decoding isolated handwritten digits from eight-channel sEMG signals. The model combines trainable signal preprocessing, convolutional feature extraction, and Transformerbased temporal modeling. It was evaluated on ten recordings from five participants using recordingseen classification, leave-one-recording-out (LORO) generalization, and few-shot adaptation. The model achieved a mean macro F1 score of 0.924 {+/-} 0.059 in the recording-seen setting and 0.619 {+/-} 0.252 under zero-shot LORO evaluation. Adaptation using two labeled trials per digit increased macro F1 to 0.828 {+/-} 0.112, while ten trials per digit achieved 0.925 {+/-} 0.053. The proposed architecture also outperformed classical and neural baselines in the controlled LORO benchmark. These results indicate that compact CNN-Transformer models, combined with lightweight target-recording calibration, provide a promising basis for adaptive sEMG-based input systems.

neuroscience↗

bioRxiv2026

Postprocessing of P300 Speller Output with a Large Language Model

P300 spellers convert electroencephalographic (EEG) activity into text by presenting users with a matrix of flickering characters. While these systems can achieve high classification accuracy, communication is severely slowed by the need for many stimulus repetitions to obtain a reliable signal. Reducing repetitions accelerates spelling but introduces character-level errors: insertions, deletions, and substitutions that degrade usability and increase user fatigue. Although substantial research has focused on improving performance at the signal acquisition and decoding stages, here we investigate a complementary text post-processing approach that leverages large language models (LLMs) to restore corrupted P300 speller output. We constructed a dataset derived from cLang-8 and simulated realistic P300-style text corruption using both random and empirically derived human-like error strategies. We evaluated several instruction-tuned LLMs alongside an optical character recognition (OCR)-fine-tuned ByT5 model under zero-shot and few-shot prompting conditions. We found that LLMs effectively recovered clean text from noisy inputs. Models employing SentencePiece tokenization consistently outperformed byte-pair encoding (BPE)-based counterparts, and few-shot in-context learning further improved restoration accuracy, with Gemma 3 achieving the strongest performance across all settings. These results suggest that LLM-based post-processing could enable P300 speller systems to operate with fewer repetitions and lower latency while maintaining or improving output accuracy, offering a practical path toward more efficient daily communication for users with motor and speech disabilities.

bioengineering↗

bioRxiv2025

Modulatory Effects of the Motor Imagery and Jendrassik Maneuver upon the Forearm H-Reflex and Electroencephalographic Activity

While the Jendrassik maneuver (JM) is commonly used in neurology to enhance monosynaptic reflexes such as the knee-jerk reflex, its interaction with cognitive processes like motor imagery (MI) remains unclear. In this study, we examined the effects of JM, MI, and their combination on the left flexor carpi radialis (FCR) H-reflex and cortical electroencephalographic (EEG) activity. Twenty healthy participants completed four tasks: resting, MI (imagining squeezing a tennis ball with the left hand), JM (squeezing a tennis ball with the right hand), and JM combined with MI (JM&MI). The FCR H-reflex was elicited with median nerve stimulation, and EEG changes were analyzed through event-related potentials and modulations of the mu rhythm. Surprisingly, JM performed with the right hand did not affect the contralateral H-reflex. MI of the left hand decreased the H-reflex in the right FCR. JM and MI significantly modulated cortical activity. Event-related desynchronization (ERD) of the mu-rhythm occurred during both JM and MI, and it was contralateral to the physical contraction or imagery. Bilateral ERD was observed during JM&MI. Cortical responses to nerve stimulation (P100, N100 and P300 components) varied by task, and during JM the P100 and P300 components correlated with strength of mu-rhythm ERD.

neuroscience↗

bioRxiv2025

Relative-location tuning in five frontal areas in a rhesus monkey

Primates exhibit remarkable flexibility in coordinating gaze angle, spatial attention, and motor goals, such as fixating on one point, attending to another, and planning a reach toward a third. Neuronal reference frames are critical for these behaviors. We investigated relative-location tuning in macaque monkeys, where the reference frame was defined by the relative positions of two objects in a scene rather than their absolute coordinates. Single neurons were recorded from five frontal cortex regions: dorsal and ventral premotor cortex (PMd and PMv), supplementary motor area (SMA), dorsolateral prefrontal cortex (PFdl), and frontal eye field (FEF). In each trial, the monkey viewed two food pellets, one in a stationary feeder and another in a feeder attached to a moving robot, which served as an attention attractor. By task rule, the monkey could obtain only one pellet. A significant proportion of neurons in all five regions showed spatial tuning based on the relative positions of the two feeders ([~]35-65%, depending on the area). For instance, some neurons were more active when one feeder was positioned to the right of the other. In PFdl and FEF, most relative-location neurons primarily encoded the monkeys fixation on the relatively left or right feeder. In SMA, PMd, and PMv, some neurons reflected fixation, but more encoded whether the relatively left or right feeder was the attention attractor (independent of gaze) or the reach target. Relative-location tuning was more prevalent when both feeders were task-relevant, such as when the monkey attended to one feeder while planning a reach to the other. Relative-location neurons could contribute to computing object-based reference frames or allocating spatial information-processing resources across attentional, oculomotor, and skeletomotor behaviors. They also suffice to explain gaze-dependent reach-related activity without contradicting theories of visually guided reaching, which predict gaze invariance.

neuroscience↗

bioRxiv2025

Neurophysiological Markers for the Acute Pain Model in Rats Under Urethane-induced Anesthesia

Certain neurophysiological mechanisms of pain can be investigated in anesthetized animals, and such a model is potentially useful for the development of pain treatment. Yet, the interference of pain-related neural patterns and anesthesia-related patterns should be understood. Here we studied the interplay between acute pain-induced and urethane-induced EEG activity. We analyzed the activity of the rats somatosensory cortex under urethane-induced anesthesia both before and after introducing a painful stimulus (formalin injection into the left hindlimb). Our analysis of the EEG parameters, such as amplitude, amplitude difference between the left and right hemispheres, and spectral entropy showed significant pain-related effects. The responses to painful stimuli depended on the anesthesia phase. These results are applicable to further research in animal models and to monitoring humans placed under general anesthesia.

neuroscience↗

bioRxiv2025

Beyond Traditional Poincare Analysis: Second-Order Plots Reveal Respiratory Effects in Heart Rate Variability

Heart rate variability (HRV) is a non-invasive biomarker of autonomic nervous system activity, commonly analyzed using a Poincare plot. This plot visualizes correlations between successive heartbeats (RRi vs. RRi+1) and quantifies autonomic regulation through SD1 and SD2 parameters. We introduce a second-order Poincare plot, a natural extension that clarifies serial dependencies by plotting successive differences in RR intervals ({Delta}RRi vs. {Delta}RRi+1). Applied to a PhysioNet dataset of 20 healthy individuals, this technique filtered out the slow HRV baseline of traditional elliptical plots to reveal distinct higher-order dynamics. These included ring-shaped structures indicating cardiorespiratory synchronization. A coupled-oscillator model, developed to simulate respiratory modulation, confirmed that these patterns are dictated by the respiratory frequency to heart rate ratio: slower breathing produces positive serial correlations in {Delta}RR, while faster breathing induces negative ones. By visualizing serial dependencies that conventional HRV metrics miss, the second-order Poincare plot extends the classical analysis framework. This tool provides a refined method for uncovering subtle dynamical features in HRV across diverse physiological and clinical states. HighlightsO_LISecond-order Poincare plots, plotting successive differences of RR intervals ({Delta}RRi vs. {Delta}RRi+1), extend traditional Poincare analysis to reveal rapid HRV dynamics. C_LIO_LIIn a dataset of 20 healthy individuals, second-order plots filtered out slow HRV components, highlighting respiratory modulation. C_LIO_LIRing-shaped patterns in some participants indicated strong cardiorespiratory coupling, while others showed positive or negative serial correlations linked to breathing rate. C_LIO_LIA coupled-oscillator model confirmed that the ratio of respiratory to heart rate frequency determines serial correlation patterns. C_LIO_LIThis method offers a novel tool for analyzing HRV dynamics, with potential applications in physiological and clinical research. C_LI

physiology↗

bioRxiv2023

The olfactory-based neurofeedback of the EEG alpha rhythm

Neurofeedback (NFB) is a form of biofeedback that enables subjects to monitor and control their own brain activity. To communicate with the subject, modern NFB training methodologies utilize various signaling mechanisms, typically visual and/or auditory stimuli. Olfaction has not been explored yet as a way to deliver NFB. Here we developed an olfactory-based NFB system based on electroencephalographic (EEG) recordings in human participants. The system incorporates an EEG recording apparatus, a custom olfactory display for the automated delivery of Sniffin Sticks, and a Python application for the conversion of EEG rhythms into NFB signals and controlling behavioral tasks. We tested occipital alpha rhythm as the source of NFB. Fifteen healthy participants were randomly assigned to three groups: olfactory neurofeedback, auditory NFB, and mock-olfactory NFB. NFB training resulted in an increase of alpha power in the true NFB groups, but not in the mock NFB group where the alpha power decreased, probably because of fatigue and drowsiness. Based on these results, we conclude that olfactory NFB is feasible, and lay out a framework for its future development.

neuroscience↗

bioRxiv2022

EEG correlates of olfactory processing during an instructed-delay task

EEG correlates of olfaction are of fundamental and practical interest for many reasons. In the field of neural technologies, olfactory-based brain-computer interfaces (BCIs) represent an approach that could be useful for neurorehabilitation of anosmia, dysosmia and hyposmia. While the idea of a BCI that decodes neural responses to different odors and/or enables odor-based neurofeedback is appealing, the results of previous EEG investigations into the olfactory domain are rather inconsistent, particularly when non-primary processing of olfactory signals is concerned. Here we report the results from an EEG study where an olfaction-based instructed-delay task was implemented. We utilized an olfactory display and a sensor of respiration to present odors in a strictly controlled fashion. Spatio-spectral EEG properties were analyzed to assess the olfactory-related components. Under these experimental conditions, EEG components representing processing of olfactory information were found around channel C4, where changes were detected for the 12-Hz spectral component. Additionally, significant differences in EEG patterns were observed when responses to odors were compared to the responses to odorless stimuli. We conclude EEG recordings are suitable for detecting active processing of odors. As such they could be integrated in a BCI that strives to rehabilitate olfactory disabilities or uses odors for hedonistic purposes.

neuroscience↗

bioRxiv2020

Generating Artificial Sensations with Spinal Cord Stimulation in Primates and Rodents

For patients who have lost sensory function due to a neurological injury such as spinal cord injury (SCI), stroke, or amputation, spinal cord stimulation (SCS) may provide a mechanism for restoring somatic sensations via an intuitive, non-visual pathway. Inspired by this vision, here we trained rhesus monkeys and rats to detect and discriminate patterns of epidural SCS. Thereafter, we constructed psychometric curves describing the relationship between different SCS parameters and the animals ability to detect SCS and/or changes in its characteristics. We found that the stimulus detection threshold decreased with higher frequency, longer pulse-width, and increasing duration of SCS. Moreover, we found that monkeys were able to discriminate temporally- and spatially-varying patterns (i.e. variations in frequency and location) of SCS delivered through multiple electrodes. Additionally, sensory discrimination of SCS-induced sensations in rats obeyed Webers law of just noticeable differences. These findings suggest that by varying SCS intensity, temporal pattern, and location different sensory experiences can be evoked. As such, we posit that SCS can provide intuitive sensory feedback in neuroprosthetic devices.

neuroscience↗

bioRxiv2019

Creating a neuroprosthesis for active tactile exploration of textures

Intracortical microstimulation (ICMS) of the primary somatosensory cortex (S1) can produce percepts that mimic somatic sensation and thus has potential as an approach to sensorize prosthetic limbs. However, it is not known whether ICMS could recreate active texture exploration--the ability to infer information about object texture by using ones fingertips to scan a surface. Here we show that ICMS of S1 can convey information about the spatial frequencies of invisible virtual gratings through a process of active tactile exploration. Two rhesus monkeys scanned pairs of visually identical screen objects with the fingertip of a hand avatar, controlled via a joystick and later via a brain-machine interface, to find the one with denser virtual gratings. The gratings consisted of evenly spaced ridges that were signaled through ICMS pulses generated when the avatars fingertip crossed each ridge. The monkeys learned to interpret these ICMS patterns evoked by the interplay of their voluntary movements and the virtual textures of each object. Discrimination accuracy across a range of grating densities followed Webers law of just-noticeable differences (JND), a finding that matches normal cutaneous sensation. Moreover, one monkey developed an active scanning strategy where avatar velocity was integrated with the ICMS pulses to interpret the texture information. We propose that this approach could equip upper-limb neuroprostheses with direct access to texture features acquired during active exploration of natural objects.

neuroscience↗

bioRxiv2019

What, if anything, is the true neurophysiological significance of rotational dynamics?

Back in 2012, Churchland and his colleagues proposed that \"rotational dynamics\", uncovered through linear transformations of multidimensional neuronal data, represent a fundamental type of neuronal population processing in a variety of organisms, from the isolated leech central nervous system to the primate motor cortex. Here, we evaluated this claim using Churchlands own data and simple simulations of neuronal responses. We observed that rotational patterns occurred in neuronal populations when (1) there was a temporal shift in peak firing rates exhibited by individual neurons, and (2) the temporal sequence of peak rates remained consistent across different experimental conditions. Provided that such a temporal order of peak firing rates existed, rotational patterns could be easily obtained using a rather arbitrary computer simulation of neural activity; modeling of any realistic properties of motor cortical responses was not needed. Additionally, arbitrary traces, such as Lissajous curves, could be easily obtained from Churchlands data with multiple linear regression. While these observations suggest that temporal sequences of neuronal responses could be visualized as rotations with various methods, we express doubt about Churchland et al.s exaggerated assessment that such rotations are related to \"an unexpected yet surprisingly simple structure in the population response\", which \"explains many of the confusing features of individual neural responses.\" Instead, we argue that their approach provides little, if any, insight on the underlying neuronal mechanisms employed by neuronal ensembles to encode motor behaviors in any species.

neuroscience↗

bioRxiv2018

On the Orientation of Entorhinal Grids

In the groundbreaking paper that eventually led to the 2014 Nobel prize in Physiology or Medicine, Hafting et al. (2005) reported that when rats forage for chocolate crumbs in a large open field, some neurons in their entorhinal cortex, called grid cells, exhibit crystalline-like responses to animal position, i.e. grids. Among several key findings documented in this article, the authors noted for the first time that the grids of different neurons can be tilted relative to each other, particularly if these neurons are far apart. In support of this claim, the researchers illustrated two neuronal subpopulations with a 7-10{degrees} difference in their grid orientations. Since these data are available online, we were able to reexamine these findings. Here we report several clarifications to the original observations of Hafting et al. First, we show that the relationship between the entorhinal grids is more complex than a single rotation: for the neuronal subpopulations analyzed by Hafting et al., one axis of the hexagonal grid is indeed tilted, but the other axes are not. Second, we show that local ensembles of entorhinal neurons are preferentially tuned to certain directions defined by the grid; this effect is unclear when single neurons are analyzed in isolation. Third, we argue that rat navigation traces are patterned instead of being random. For example, the orientation of the vector field representing average velocity appears to match the orientation of the neuronal grid. Overall, our observations indicate that additional insights into the function of entorhinal grids could be provided by ensemble-level analyses and thorough examination of the connection between the navigation behavior and neuronal patterns.\n\nHighlightsO_LIWhile our examination of the online dataset from Hafting et al. generally confirms their original findings, several clarifications should be made.\nC_LIO_LIFor the two neuronal subpopulations, where Hafting et al. reported a 7-10{degrees} relative tilt between the grids, only one of the grid axes is tilted, whereas the others are not.\nC_LIO_LIWhen spatial response fields are plotted for neuronal subpopulations instead of single neurons, it is clear that each subpopulation exhibits spatially periodic bands aligned with one of the grid axes.\nC_LIO_LINavigation traces are not random and appear to match the orientation and periodicity of the neuronal grid.\nC_LI

neuroscience↗

bioRxiv2017

Pitfalls in the Assessment of Brain-Machine Interfaces Using Information Transfer Rate

Information transfer rate (ITR), measured in bits/s, can be applied to evaluate motor performance, including the capacity of brain-machine interfaces (BMIs) to control external actuators. In a 2013 article entitled \"Transfer of information by BMI\" and published in Neuroscience, Tehovnik and his colleagues utilized ITR to assess the performance of several BMIs reported in the literature. We examined these analyses closely and found several fundamental flaws in their evaluation of ITR. Here we discuss the pitfalls in Tehovniks measurements of ITR, as well as several other issues raised in \"Transfer of information by BMI\", including the claim that BMIs cannot be a reasonable option for paralyzed patients.\n\nHighlightsO_LIInformation transfer rate is discussed for BMI experiments, where subjects reach to targets.\nC_LIO_LITask settings, not just the number of possible targets, are important to calculate information correctly.\nC_LIO_LIActive tactile exploration can be quantified as information transfer, but the number of targets is insufficient for such quantification.\nC_LIO_LIInformation transfer rate increases with the number of neural recording channels.\nC_LIO_LIFor practical applications, improvement in quality of life is essential, not information transfer rate per se.\nC_LI

neuroscience↗