bioRxiv Subject Collection: Neuroscience's Journal
[Most Recent Entries]
[Calendar View]
Saturday, September 27th, 2025
| Time |
Event |
| 1:33a |
Deep Learning-Based Control of Electrically Evoked Activity in Human Visual Cortex
Visual cortical prostheses offer a promising path to sight restoration, but current systems elicit crude, variable percepts and rely on manual electrode-by-electrode calibration that does not scale. This work introduces an automated data-driven neural control method for a visual neuroprosthesis using a deep learning framework to generate optimal multi-electrode stimulation patterns that evoke targeted neural responses. Using a 96-channel Utah electrode array implanted in the occipital cortex of a blind participant, we trained a deep neural network to predict single-trial evoked responses. The network was used in two complementary control strategies: a learned inverse network for real-time stimulation synthesis and a gradient-based optimizer for precise targeting of desired neural responses. Both approaches significantly outperformed conventional methods in controlling neural activity, required lower stimulation currents, and adapted stimulation parameters to resting state data, reliably evoking more stable percepts. Crucially, recorded neural responses better predicted perceptual outcomes than stimulation parameters alone, underscoring the value of our neural population control framework. This work demonstrates the feasibility of data-driven neural control in a human implant and offers a foundation for next-generation, model-driven neuroprosthetic systems, capable of enhancing sensory restoration across a range of clinical applications. | | 10:46a |
Subthreshold violations of trajectory predictions are sensitive to TMS of Cerebellum CRUS I/II
Temporal prediction can help to follow a trajectory. In case of an error, the prediction can be adjusted. However, processing the error and adjusting the prediction can take time. What happens immediately after a prediction error, and can the processing of the prediction be modulated? We use a newly found illusion based on moving squares and requiring trajectory regularity to be elicited. We examined the conscious consequences of a sub-threshold manipulation of the square trajectories, and transcranial magnetic stimulation (TMS) on the cerebellum (right CRUS I/II) to study the modulation of the processing of the trajectory manipulations. The TMS was a typical intermittent thetaburst stimulation, but only one sequence of around 3 minutes, compared with a placebo stimulation. The trajectory manipulation had a reliable effect on the illusion, even though the illusion emerged within less than 100 ms after the trajectory manipulation. The results suggest that the prediction is temporarily stopped after the trajectory change. The illusion was accompanied by EEG signals whose amplitude was modulated by TMS on the cerebellum, at least in those participants who received verum TMS after having performed the task three times. Those EEG signals resembled a late LPP (Late Positive Potential). As LPP spontaneously decreased over time, the results suggest the effect of TMS may represent a reinstation of the EEG consequences of the prediction error, i.e., a modulation of its significance. | | 10:46a |
GluN2A-mediated currents and calcium signal in human iPSC-derived neurons
Gene expression data indicate that during human brain development, neurons change the NMDA receptor (NMDAR) subunit composition to modulate their function, favouring the GluN2A subunit over GluN2B - a hallmark of neuronal maturation. However, evidence supporting this phenomenon in human iPSC-derived neurons remains elusive. Here, using two differentiation methods in parallel (BrainPhys Neuronal Medium, BPM, and Neural Maintenance Medium, NMM), we provide evidence of increased synaptic localization of NMDARs during neuronal maturation and that GluN2A subunit is crucial for the NMDA physiological function-inducing inward currents and calcium entrance at 60 days of differentiation. Calcium responses to specific agonists, particularly NMDA, were elevated in cells cultured under BPM conditions. This is likely attributable to their more mature neuronal phenotype and the RNA-seq-identified upregulation of genes involved in intracellular calcium signaling proteins. Our results offer insight into how glutamate receptor subunits mature during brain development, delineating approaches to study NMDAR activity in health and disease. | | 10:46a |
Task-specific theta enhancement and domain-general alpha/beta suppression as oscillatory signatures of individual differences in cognitive flexibility
Cognitive flexibility - the ability to adapt to changing situational or task demands - is a fundamental aspect of human cognition and a potential source of individual differences in higher-order cognitive abilities. To further understand the neural mechanisms underlying this ability, we examined event-related spectral perturbations (ERSP) in three cued switching tasks (parity/magnitude, global/local, number/letter) in a sample of 148 adults. Employing a data-driven approach that combined mass-univariate time-frequency analysis, cluster-based permutation testing, and latent change structural equation modeling, we identified two robust neural signatures: a transient frontal theta increase and a sustained alpha/beta suppression. Both effects were more pronounced in switch compared to repeat trials. Frontal theta reflected task-specific control processes, whereas parietal alpha/beta indexed domain-general attentional processes that persisted from cue onset through target processing -consistent with two stage-models of task switching. Notably, flexibility-related alpha/beta dynamics did not correlate with measures of intelligence or working memory capacity (WMC), underscoring the distinctiveness of cognitive flexibility from general cognitive abilities. These findings provide the first direct evidence that alpha/beta suppression constitutes a reliable, generalizable neural signature of individual differences in cognitive flexibility. | | 10:46a |
Visual contributions to the perception of speech in noise
Investigations of the role of audiovisual integration in speech-in-noise perception have largely focused on the benefits provided by lipreading cues. Nonetheless, audiovisual temporal coherence can offer a complementary advantage in auditory selective attention tasks. We developed an audiovisual speech-in-noise test to assess the benefit of visually conveyed phonetic information and visual contributions to auditory streaming. The test was a video version of the Children's Coordinate Response Measure with a noun as the second keyword (vCCRMn). The vCCRMn allowed us to measure speech reception thresholds in the presence of two competing talkers under three visual conditions: a full naturalistic video (AV), a video which was interrupted during the target word presentation (Inter), thus, providing no lipreading cues, and a static image of a talker with audio only (A). In each case, the video/image could display either the target talker, or one of the two competing maskers. We assessed speech reception thresholds in each visual condition in 37 young ([≤]35 years old) normal-hearing participants. Lipreading ability was independently assessed with the Test of Adult Speechreading (TAS). Results showed that both target-coherent AV and Inter visual conditions offer participants a listening benefit over the static image audio-only condition, with the full AV target-coherent condition providing the most benefit. Lipreading ability correlated with the audiovisual benefit shown in the full AV target-coherent condition, but not the benefit in the Inter target-coherent condition. Together our results are consistent with visual information providing independent benefits to listening, through lip reading and enhanced auditory streaming. | | 10:46a |
Predictive learning enables compositional representations
The brain builds predictive models to plan future actions. These models generalize remarkably well to new environments, but it is unclear how neural circuits acquire this flexibility. Here, we show that compositional representations emerge in Recurrent Neural Networks (RNNs) trained solely to predict future sensory inputs. These representations have been observed in different areas of the brain, for example, in the motor cortex of monkeys, which have been shown to reuse primitives in sequences. They enable compositional generalization, a mechanism that could explain the brain's adaptability, where independent modules representing different parts of the environment can be selected according to context. We trained an RNN to predict future frames in a visual environment defined by independent latent factors and their corresponding dynamics. We found that the network learned to solve this task by developing a compositional internal model. Specifically, it had disentangled representations of the static latent factors, and formed distinct, modular clusters, each selectively implementing a single dynamic. This modular and disentangled architecture enabled the network to exhibit compositional generalization, accurately predicting outcomes in novel contexts composed of unseen combinations of dynamics. Our findings present a powerful, unsupervised mechanism for learning the causal structure of an environment, suggesting that predicting the future can be sufficient to develop generalizable world models. | | 7:31p |
Beyond Agreement: Standardizing Crowdsourced Synapse Annotations through Proofreading in EM Connectomics
Reliable synapse identification in volumetric EM is hampered by subtle, 3D cues that yield variable human judgments. We present a standardized proofreading protocol that pairs explicit, operational criteria with machine-learning candidate generation and a two-stage calibration of annotators. In two larval Drosophila melanogaster volumes imaged at 8x8x8 nm, five raters (expert + 4 calibrated annotators) reviewed model-proposed candidates using efficient node-based labels. Multi-rater judgments were aggregated with a probabilistic Dawid-Skene (DS) model to produce consensus labels with calibrated uncertainty. Post-calibration, individual annotator accuracy versus the expert improved (McNemar p<0.05 for all raters), DS-expert agreement increased, and DS posterior entropy decreased for true positives/negatives, indicating more decisive consensus; gains were modest and dataset-dependent in chance-corrected agreement (Krippendorff's ). By making uncertainty explicit, this protocol converts noisy judgments into auditable supervision suitable for training and evaluation, while honestly communicating residual ambiguity essential for reliable and robust connectomics at scale. |
|