NEUROLYMPICS Conducts First Full-Scale Data-Capture Session with 20 Test Participants

Abrantes, Portugal

The NEUROLYMPICS consortium held its first large-scale data-capture and demonstration session at the project's facilities in Abrantes, bringing together representatives from all three co-promoters: Techframe, Neroes and the University of Leiria and Oeste (ULO), to observe and support twenty volunteer participants as they operated a brain-computer interface (BCI) prototype developed under the project.

The session forms part of Activity 5 of the NEUROLYMPICS work plan, dedicated to prototype construction, testing and trials, and represents a first structured contribution to the large-scale neurophysiological reference database the consortium is building to train and validate its machine-learning models, one of the project's central technical objectives, alongside a target F1 classification score of 95% and calibration times short enough for everyday use.

Each participant was fitted with the project's EEG headset and complementary physiological sensors before undergoing a brief calibration routine, during which the system established an individual baseline for each subject's brain activity. Participants were then asked to complete a control task using only their mental state: lifting a virtual spacecraft off the ground and sustaining stable flight for as long as possible, a scenario designed to test both the responsiveness and the precision of the underlying signal-classification pipeline. 

Members of the Neroes team led the physiological signal acquisition and neurofeedback protocol, drawing on the company's experience with its Mental Training Platform, while the Techframe team oversaw the real-time processing pipeline and the interactive engine driving the spacecraft demonstration. Researchers from ULO's Centro de Investigação em Informática e Comunicações (CIIC) monitored classification performance throughout the session, cross-referencing live results against the project's target metrics for accuracy, response time and command discrimination. 

Results across the twenty sessions varied considerably, offering the technical team a valuable spread of data. Several participants needed a few attempts before achieving lift-off, describing an initial tendency to "try too hard", consciously willing the spacecraft upward rather than sustaining the calm, focused mental state the system responds to best. Once that adjustment was made, most were able to achieve stable lift within the session. Maintaining steady flight proved to be the more demanding half of the task for many participants: momentary lapses in concentration were consistently reflected in the spacecraft's trajectory, illustrating in real time the sensitivity of the classification model to sustained attentional states. A smaller subset of participants adapted markedly faster than the group average, progressing through the demonstration's difficulty levels within minutes of their first successful attempt. 

For the NEUROLYMPICS team, this variability is itself a valuable outcome. Every session, whether it ends in a smooth, sustained flight or a series of short, unstable attempts, adds another labelled data point to the reference dataset underpinning Research Area CI1, while also generating direct user-experience input for the adaptive game design work under CI5 and ergonomic feedback relevant to the physical BCI device under development in CI6. 

The consortium plans to run further data-capture sessions of this kind at regular intervals over the coming months, progressively expanding the reference database and refining the system's classification models as the project advances toward its mid-term milestones.