A computational model of the brain's thalamo-cortico-thalamic network, built from twelve neural populations, reproduces the abnormal EEG oscillations seen in schizophrenia by simulating the effects of ketamine, which blocks NMDAR receptors. The model shows that ketamine increases excitatory activity and alters gamma and sigma band oscillations, matching experimental observations. Adding a neuroplasticity model of transcranial direct current stimulation (tDCS) and applying simulated current to selected pathways reverses these ketamine-induced changes. The work suggests that neural mass models can help predict personalized tDCS protocols for treating schizophrenia.
The brain's spontaneous activity has intrinsic durations—Intrinsic Neural Timescales (INTs)—that are hierarchically organized, with shorter durations in sensory regions and longer ones in association areas. This study shows that the topographic organization of INTs is not fixed but dynamically changes over time. Healthy individuals exhibit transitions between different INT states that are moderately predictable and show memory effects. In people with disorders of consciousness, these transitions become less predictable and show reduced memory effects, suggesting that the temporal richness of INT state transitions is important for maintaining normal consciousness.
A computational model of an extended corticothalamic network simulates how different concentrations of propofol alter states of consciousness. The model shows that oscillation spread across the cortex moves from posterior to anterior brain regions in four sequential stages, driven by differences in connections between regions. This suggests that hierarchical rhythm propagation depends on the heterogeneity of inter-regional connectivity. The model provides a millisecond-resolution simulation platform for studying how brain areas coordinate activity during consciousness and how anesthetics affect brain regions.
Using physiological data from 25 subjects wearing the Empatica E4 wristband, Emotiv Epoc X EEG headset, and Biosignalplux RespiBAN during arithmetic and reading tasks, feature engineering and deep feature learning approaches automatically discriminated between flow and non-flow states. EEG sensor modalities alone achieved 64.97% accuracy and a macro averaged F1 score of 64.95%; fusing all sensor modalities yielded 73.63% accuracy and an AF1 of 72.70%. A transfer learning approach using emotional arousal classification on the DEAP dataset improved performance to 75.10% accuracy and an AF1 of 74.92%. The results suggest effective discrimination is possible with multimodal sensor data, and the success of transfer learning indicates that emotions and flow are connected.