Student Speakers 2025

Toward Gaze-independent BCI Spellers: Decoding covert Attention
by Radovan Vodila, Jordy Thielen

Presenter: Radovan Vodila

A limitation of brain-computer interface (BCI) spellers is that they require the user to be able to move the eyes to fixate on targets. This poses an issue for users who cannot voluntarily control their eye movements, for instance, people living with late-stage amyotrophic lateral sclerosis (ALS). This study makes the first step towards a gaze-independent speller based on the code-modulated visual evoked potential (c-VEP). Participants were presented with two bi-laterally located  stimuli, one of which was flashing, and were tasked to covertly attend to one of these stimuli by using spatial attention, eliminating the need for eye movement. Electroencephalography (EEG) recordings were analyzed to extract three neural signatures: the c-VEP signal, lateralization of alpha-band oscillations, and ERP components associated with target detection (P300). Classification accuracies for decoding the attended side were approximately 88% for c-VEP, 85% for alpha-band lateralization, and 96% for ERP responses, demonstrating the individual potentials of these markers. These fundamental insights showcase the promising feasibility of gaze-independent BCIs that use covert spatial attention when both stimuli flash simultaneously.

Maintenance and Manipulation of Retrieved Long-Term Memories in Working Memory
by Giulia Quagliozzi, prof. Elkan Akyürek

Presenter: Giulia Quagliozzi

Memories dynamically shape ongoing cognition, yet how memories are maintained in an accessible state following their retrieval remains largely unexplored. This study investigates the fate of long-term memories following their cortical reinstatement, including how they are altered within working memory to meet new task demands. To do this, participants learned sound-orientation pair associations and retrieved these orientations from long-term memory for a consequent mental rotation task, while electroencephalography (EEG) was used to monitor neural activity. During the delay periods following memory reactivation and manipulation, two neutral “impulse” stimuli were administered to elicit stimulus-related neural activity, facilitating detection in cases where activity may have subsided. We hypothesized that retrieved memories would be retained in working memory to ensure their availability for subsequent manipulation, evidenced by the presence of stimulus- specific neural activity in the period following their reactivation (Hypothesis 1). Additionally, we expected both original and rotated memory representations to be simultaneously maintained in working memory during the mental rotation task (Hypothesis 2). Based on multivariate pattern analysis with Mahalanobis distance classifiers, the results showed that memory orientations were consistently decodable throughout the trial, with peaks in accuracy during reinstatement and manipulation, indicating that retrieved memories were actively sustained in working memory. No increase in decoding accuracy was observed following impulse presentations, suggesting reliance on continuous neural spiking rather than activity- quiescent states. Surprisingly, rotated representations were not decodable, highlighting potential methodological limitations or task complexity. In conclusion, the study deepens our understanding of the neural mechanisms governing the interaction between retrieved long-term memories and working memory, offering insights into the processes of memory maintenance and manipulation within the cortex.

Burst-Based Plasticity in CA1 Neurons: A Computational Model of Place Cell Formation and Remapping
by Felix Dörr, Federico Stella

Presenter: Felix Dörr

Hippocampal place cells in CA1 form through input from both CA3 and the Entorhinal Cortex (EC). CA3 provides direct input to basal and oblique dendrites, supporting memory retrieval and sequence generation, while EC projects to apical dendrites, conveying high-level spatial information. When EC input induces plateau potentials coinciding with CA3-driven somatic activity, it triggers bursts, which play a key role in synaptic plasticity and learning. Experiments show that place cells emerge through spontaneous bursts, stabilizing to encode spatial locations. Bursting has been linked to effective credit assignment in neural circuits, shaping learning dynamics. Moreover, experimental studies have shown that silencing EC-CA1 connections alters spatial remapping, but the underlying mechanisms remain unclear. We use a computational model to investigate how EC input influences map stability and remapping in novel environments. We use a computational model of hippocampal CA1 consisting of two-compartment leaky integrate-and-fire neurons capable of both spiking and bursting. The network includes two types of interneurons to regulate dynamics, one specifically inhibiting the basal compartment and one inhibiting the apical compartment. The neurons receive direct spatial input from the Entorhinal Cortex (EC) at their apical dendrites, while spatial input from CA3 arrives via a connection matrix representing the Schaffer Collaterals. Synaptic plasticity of this connection matrix follows a learning rule depending both on bursts and on Hebbian Plasticity. We simulate an experimental paradigm in which EC-CA1 connections are optogenetically silenced in a subset of conditions. The protocol consists of five sequential phases: (1) familiarization with a known environment (F1, EC on), (2) recall with EC silenced (F2), (3) exposure to a novel environment (N1, EC off), (4) return to the familiar environment (F3, EC on), and (5) re-exposure to the novel environment (N2, EC on). A control group experiences the same sequence but retains EC input throughout. Our model reproduces key experimental findings. In the experimental group, N1 fails to induce remapping, while the control group exhibits robust remapping. Strikingly, even in N2, remapping does not occur in the experimental group, suggesting that the familiar representation from F1 persists despite novel sensory input. This effect arises due to pattern completion via Schaffer collateral connections between CA3 and CA1, which reinforce the pre-existing map in the absence of direct EC input. This is supported by higher correlation between the spatial maps of the two environments in CA1 than in CA3. These results provide computational evidence that EC input is essential for spatial remapping. When EC is silenced, CA3-driven pattern completion dominates, leading to rigid map stability even in novel environments. Moreover, it suggest that such a learning signal can be encoded in bursts as supported by prior experimental and theoretical results. This suggests a crucial role for EC in enabling flexible spatial representations in the hippocampus.

Neuronal Activity in the Prefrontal Cortex in Susceptibility and Resilience to Stress
by Emanuela Pirani, Ana Carolina Temporão, Marloes Henckens

Presenter: Emanuela Pirani

Post-traumatic stress disorder (PTSD) is a severe anxiety disorder that develops following exposure to traumatic events and is characterised by symptoms such as re-experiencing, avoidance, negative emotions, and hyperarousal. These symptoms can persist for extended periods and impact physical, psychological, professional, and social domains. Stress resilience, the ability to maintain mental health despite adversity, is a crucial factor distinguishing individuals who develop PTSD from those who do not. At the cellular level, trauma memory processing involves engrams, stable yet flexible neuronal assemblies associated with memory encoding and recall. The dynamic nature of engram formation and drift raises questions about how stable memories are maintained despite neuronal variability. This study aims to identify neural mechanisms that differentiate PTSD-susceptible individuals from resilient ones by investigating trauma memory processing in a mouse model. Specifically, we focus on the role of the prefrontal cortex (PFC) in regulating adaptive versus maladaptive trauma memories. A stress-enhanced fear learning (SEFL) paradigm was utilised to induce PTSD-like stress responses in mice, creating susceptible and resilient groups. Behavioural tests measured hypervigilance, anxiety, risk assessment, and sleep disturbances. To investigate trauma memory encoding, Arc+ neurons were labelled in ArcTRAP mice following stress exposure. Remote recall was assessed three weeks later using c-Fos immunohistochemistry to detect neural activation in the PFC after re-exposure. This study contributes to the understanding of resilience by identifying neural circuits associated with adaptive trauma responses. These findings have potential implications for developing targeted interventions to promote resilience in individuals at risk of PTSD.