Decoding local and global fNIRS patterns of kinesthetic and visual-motor imagery in a haptic interface paradigm

Motor imagery (MI) involves visual-motor (VMI) or kinesthetic-motor (KMI) perspective-taking. However, there is limited knowledge of the accuracy of identifying the subject's perspective during MI using brain activity patterns. The aim of this study is to use a novel haptic interface paradigm to classify cerebral hemodynamic responses obtained via functional near-infrared spectroscopy (fNIRS) under KMI/VMI and REST conditions. The study included 29 right-handed participants (17 females) aged 18–46 (M = 23, SD = 6.32). Significant differences in classification results were observed between VMI and REST in the frontopolar area, the left supplementary motor cortex, and the visual cortex. Furthermore, we demonstrated the ability to discriminate between KMI and REST (M = 83%, SD = 11%; α = 1%) and between VMI and REST (M = 71%, SD = 12%; α = 5%). Additionally, among the 120 identified signal-processing pipelines, we selected those that achieved the best classification performance across both local (selected regions of interest) and global (all channels) hemodynamic patterns. These results suggest that this haptic interface paradigm could be useful in future neurofeedback training (NFT) or brain-computer interface (BCI) applications for clinical groups.