Next Gen: Trainee Spotlights (free event)

Presentation title: Multi-scale dynamical modeling for spike and LFP activity using a NeuroBondGraph Network

Abstract: The brain exhibits computational structure across a variety of scales: from single neurons (micro-scale) to functional areas (meso-scale) to cortical networks (macro-scale). In neuropsychiatric conditions, pathological activity is typically neural circuit-wide, impacting dynamics at multiple scales, either directly or indirectly. Combining information on changes in brain dynamics at multiple scales can shed new light on the biological cause of neuropsychiatric disorders.. While single-level neural dynamics encode rich information giving rise to behavior, the different levels of brain network can communicate with each other. However, how the activity and information at one level interact with other levels is seldom explored. Evidenced by the success of cross-scale modeling between local field potentials (LFP) and signals recorded from screw type electrodes implanted in the skull via the NeuroBondGraph Network (NBGNet) in the previous work, we develop the multi-scale NBGNet to study neural communications between spike and LFP in the brain networks. Neural representation extracted from multi-scale NBGNet can potentially explain the naturalistic behaviors, which will be validated in real-time brain-machine interface experiments. With the incorporation of realistic brain constraints in the multi-scale NBGNet, this neurobiologically realistic framework holds great potentials to improving our understanding of complex brain functions in unprecedented detail.

Biography: I am a Ph.D. student in Santacruz lab since Fall 2019. I earned a M.Sc. degree in Mechanical Engineering from the University of Texas at Austin in 2018. My research focuses on the development of the multi-scale dynamical model to integrate multi-modal brain signals and to illuminate the mechanistic understanding of brain computation.

Presentation Title: Influence of feedback transparency on performance in beta-based motor imagery neurofeedback

Abstract: Our study focuses on Motor-Imagery BCI (MI-BCI) feedback design. Standard MI-BCI protocols feature abstract feedback (e.g., visual gauge) unrelated to the imagined movement. Transparent feedback would be directly related to the task. This may reinforce agency—the ability to self-attribute actions, such as feedback movements—leading to lessened complexity in BCI control, and ultimately improving BCI performance. We tested this hypothesis in a beta-based MI neurofeedback task with electroencephalography (EEG). Twenty-three subjects participated in a single right-hand MI neurofeedback session. Feedback was based on online laplacian-filtered 8-30Hz power on C3 electrode computed with OpenViBE. There were 3 feedback conditions of assumed increasing transparency: pendulum; virtual hand; virtual hand with additional motor illusion vibrations to the participant’s hand. Participants performed two runs of five 24s-trials per condition. Agency and user experience questionnaires were included. Repeated-measures ANOVA on the median event-related desynchronisation (ERD) values computed from feedback logs showed greater BCI performance for virtual hand than pendulum feedback condition, but lower performance for visuotactile than virtual hand-only feedback. Agency showed a similar pattern. Feeling of success did not vary with feedback conditions. BCI performance was correlated with agency. Thus, visual transparency—but not tactile feedback addition—increased BCI performance and this was related to agency.

Biography: I graduated from the Bordeaux Graduate School of Cognitive Engineering in 2020 and pursued a MSc degree in Cognitive Science in 2021. After two experiences in neurofeedback research, I decided to carry on and start a PhD at the Paris Brain Institute. My thesis project focuses on the specificity and selectivity of MI-based EEG neurofeedback. I investigate the impact of task-feedback transparency and operant learning on user training and neurofeedback performance. The final purpose of my studies is the development of an adapted, optimized neurofeedback protocol for patients with Parkinson’s disease.

Presentation Title: Channel selection for Low-Density EEG source reconstruction

Abstract: Source reconstruction is the estimation of the neural activity inside the brain form the EEG signals recorded at the scalp. High-density EEG (HD-EEG) is proven to be the most accurate option, where higher number of channels offer a higher spatial resolution that allows to discriminate better the brain sources. However, in multiple EEG applications, in which BCI systems is one of the most relevant, the use of HD-EEG systems can be inconvenient due to practical and portability issues, and the increased volume of information to process. We propose a channel selection methodology that identify subsets of channels that can accurately reconstruct a set of sources of a particular brain activity. We have evaluated this by using a forward model, and a simulated ground-truth set of 150 EEG trials with three active sources with temporal mixing. Preliminary results suggest that with 8 channels is possible to obtain a similar accuracy as with 231 ch., in which for the 60% of the trials an equal or better accuracy was obtained with the selected channels. This can impact the design of multiple low-density EEG applications in which is required to extract the cortical activity of specific brain regions.

Biography: I am a PhD candidate of Norwegian University of Science and Technology; I hold a M.Sc. degree in Electrical engineering from Technological University of Pereira. My research focus are the EEG-based brain imaging methods, and EEG signal processing and analysis. I am particularly interested on studying signal processing methods for portable EEG applications, and the estimation of the brain source activity with low-density EEG systems. Other current research interests include the application of AI methods for classifying the neural activity using cortical reconstructed activity, identification of biomarkers from EEG signals, and the application of metaheuristic optimization techniques in EEG research.