The underlying mechanisms of tDCS-Based cognitive enhancement in unpredictable task switching: A machine learning-based approach for EEG signal analysis.

Transcranial direct current stimulation (tDCS) enhances cognitive abilities yet has highly inconsistent outcomes, highlighting the need to clarify its neurophysiological mechanisms. Herein, we integrated EEG with machine learning to assess 24 participants performing unpredictable magnitude/parity switching tasks under anodal tDCS (a-tDCS) and sham conditions using a double-blind design. Behavioral results showed that a-tDCS over the right dorsolateral prefrontal cortex (rDLPFC) markedly improved switch cost accuracy while reducing mixing cost accuracy. EEG analyses revealed that a-tDCS induced a significantly larger N1 in repeat trials, eliminated switch-repeat differences in N2 latency, frontal/central alpha (1200-1500 ms) and parietal beta power (600-800 ms), while enhancing central theta power (100-400 ms) in females. Moreover, participants were divided into high-gain and low-gain groups by median split of switch cost improvement score (a-tDCS minus sham accuracy). Decoding analyses further revealed a significantly higher trial-type classification accuracy in the low-gain group under a-tDCS condition during stimulus-response interval. Conversely, the high-gain group exhibited lower trial-type classification accuracy, albeit with a higher response-hand classification accuracy during pre-stimulus period. Additionally, model interpretability analyses showed a frequency band main effect on trial-type classification with significantly higher alpha and/or beta weights compared to theta and/or delta weights across regions. Furthermore, a-tDCS significantly reduced left frontal delta weights, enhanced left centrofrontal alpha weights, and increased right central alpha weights in high-gain group. These findings suggest that a-tDCS over the rDLPFC regulates attention, conflict detection, and inhibition to modulate unpredictable task switching, highlighting the dominant role of high-frequency bands in trial-type classification, and indicating that a-tDCS impacts the low- and high-gain groups via reactive conflict resolution and broad proactive preparation, respectively.