Figure 1 Integrated framework of BCI-based neurorehabilitation for post-stroke psychiatric disorders. This diagram illustrates the closed-loop mechanism of BCI intervention through 3 interconnected stages: (Left) Multimodal signal acquisition: real-time capture of EEG rhythms and fNIRS hemodynamic responses from the patient, specifically monitoring activity around the stroke lesion. (Middle) AI-driven signal decoding: raw neural data is processed by deep learning algorithms to accurately classify mental states and generate neuromodulation parameters. (Right) Prefrontal-limbic circuitry regulation: the intervention targets key emotional and cognitive regulation centers, including the dorsolateral prefrontal cortex, anterior cingulate cortex, amygdala, and hippocampus. (Bottom right bubble) Synaptic plasticity: at the cellular level, the feedback loop promotes long-term potentiation (LTP), dendritic spine growth, and neurotransmitter regulation, ultimately leading to reduced depression/anxiety and improved cognitive function. BCI – brain computer interface; EEG – electroencephalography; fNIRS – functional near-infrared spectroscopy; AI – artificial intelligence; LSTM/CNN – long short-term memory/convolutional neural network; LTP – long-term potentiation.