The influence of cognitive activity on subsequent daytime nap: A deep neural network model based on sleep spindles

•The influence of cognitive activity avoiding emotion on sleep is investigated.•No differences in macro-sleep and sleep spindles characteristics between conditions.•Deep neural network model achieved 96% accuracy in differentiating conditions.•Our findings suggest a weak but positive effect of cognitive activity on sleep. Understanding the influence of cognitive activity on subsequent sleep has both theoretical and applied implications. This study aims to investigate the effect of pre-sleep cognitive activity, in the context of avoiding emotional interference, on macro-sleep and sleep spindles. In a within-subjects design, participants’ sleep electroencephalography was collected in both the with and without pre-sleep cognitive activity conditions. Subsequent macro-sleep (i.e., sleep stage distribution and sleep parameters) and spindle characteristics (i.e., density, amplitude, duration, and frequency) were analyzed. In addition, a novel machine learning framework (i.e., deep neural network, DNN) was used to discriminate between cognitive activity and control conditions. There were no significant differences in macro-sleep and sleep spindles between the cognitive activity and control conditions. Spindles-based DNN models achieved over 96% accuracy in differentiating between the two conditions, with fast spindles performing better than full-range and slow spindles. These results suggest a weak but positive effect of pre-sleep cognitive activity on subsequent sleep. It sheds light on a possible low-cost and easily accessible sleep intervention strategy for clinical and educational purposes.