Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session A: Tuesday, August 12, 1:30 – 4:30 pm, de Brug & E‑Hall

What are the best features to decode the levels of working memory load from ECoG data?

Funda Yilmaz¹, Seyma Nur Ertekin²; ¹Donders Institute for Brain, Cognition and Behaviour, ²University of Amsterdam

Presenter: Funda Yilmaz

This study aims to decode the three levels of working memory load in n-back task (0-back, 1-back and 2-back) from ECoG data, utilising different feature selection strategies using regularised logistic regression. The results demonstrated that feature strategies based on common electrodes across subjects yielded the highest classification accuracies within individualized models, followed by selection of specific brain regions, combined with data-driven methods. We also employed time-frequency analysis to differentiate potential neural markers. The results showed that low-frequency oscillations carried the most discriminative information. Furthermore, our findings indicate that the neural signature of working memory load varies between participants, yet certain cross-participant features appear to be conserved. Overall, effective feature selection may enhance both the interpretation of workload-related neural activity and the performance of simple algorithms.

Topic Area: Memory, Spatial Cognition & Skill Learning

Extended Abstract: Full Text PDF