Electroencephalogram-Based Sustained Attention Assessment Using Sparse Model for Feature Selection

Sustained attention, or concentration, refers to the capacity to remain focused on a particular task over an extended period. Concentration assessment has traditionally relied on subjective questionnaires and behavioral measures such as response time to targeted stimuli, yet electroencephalography (EEG) offers a promising neurophysiological approach to quantify attentional states. In this study, we employed a go/no-go paradigm while recording EEG signals and behavioral response times to examine the neural correlates of concentration. To identify the most informative neurophysiological markers, we implemented group Lasso feature selection on a comprehensive set of EEG parameters, including relative spectral powers, power ratios, and entropy measures across multiple electrode sites. This approach enables systematic identification of the most predictive combinations of spatial and spectral EEG features for assessing attentional states. Our results show that the frontal, right parietal, and occipital electrodes and their power ratios are the most effective variables for assessing concentration. The results also reveal that concentration can be assessed using a small subset of electrodes and features without significantly affecting performance.