The Cortical Spatial Responses and Decoding of Emotion Imagery Toward a Novel fNIRS-Based Affective BCI.

Functional near-infrared spectroscopy (fNIRS), with its non-invasive and high spatial resolution, holds promise in developing novel affective brain-computer interface (BCI). Similar to motor imagery BCI, emotion imagery BCI could recognize internal emotions and convey them to the external world. This holds clinical value for expressing emotions in patients with neurological impairments and serves as a proactive emotion regulation method. However, the fNIRS features of emotion imagery for affective BCI and the discriminability of different emotion categories remain unclear. Here, this study designed a novel emotion verbal imagery paradigm (imagining descriptions of happy or sad scenes). First, task-related hemodynamic responses were analyzed from 17 subjects. Then, statistical analyses were then conducted to reveal the significant cortical spatial response patterns. Additionally, decoding experiments and model interpretability are employed to assist in validating the feasibility of the emotion imagery BCI. Results showed: 1) Happy imagery recruited frontoparietal regions, such as the left dorsal secondary motor cortex, ventral secondary motor cortex, and inferior parietal lobe. 2) Sad imagery mainly recruited the right dorsolateral prefrontal cortex. 3) The left dorsal sensorimotor cortex exhibited selective responsiveness to happy imagery and sad imagery. 4) The classification results of the emotion imagery task exceeded the random level. 5) Emotional categories activation responses showed significant similarity with the hemodynamic responses of the imagination tasks. Taken together, by proposing the emotion imagery fNIRS paradigm, this work could shed light on the development of feature non-invasive BCI.