Objective : This study reveals the changes in brain activation due to different game interaction states based on functional near-infrared spectroscopy signals and discusses their significance for stroke rehabilitation. Approach : The oxygenated hemoglobin concentration (Delta [HbO ₂ ]) signals and the deoxygenated hemoglobin (Delta [HbR]) signals were recorded from the prefrontal cortex (PFC), the motor cortex (MC), the occipital lobe (OL) and the temporal lobe of 21 subjects (mean age: 24.6 ± 1.9 years old) in three game interaction states: physical, motion-sensing, and button-push training. The subjects were also asked to complete user-satisfaction survey scales after the experiment. Main results : Compared with the button-training state, several channels in the PFC and MC region of the physical-training state were significantly altered as were several channels in the RMC region of the motion-sensing training state ( P < 0.05 after adjustment). The motion-sensing state of the PFC had a significant correlation with that of the MC and the OL. The subjective scale results show that the acceptability of the physical and motion-sensing states was greater than the acceptability of the button-push training state. Significance : The results show that the brain regions responded more strongly when activated by the physical and motion-sensing states compared with the button-push training state, and the physical and motion-sensing states are more conducive to the rehabilitation of the nervous system. The design of rehabilitation products for stroke patients is discussed and valuable insights are offered to support the selection of better interactive training methods.

In Industry 4.0, medical data present a trend of multisource development. However, in complex information networks, an information gap often exists in data exchange between doctors and patients. In the case of diseases with complex manifestations, doctors often perform qualitative analysis, which is macroscopic and fuzzy, to present treatment recommendations for patients. Improving the reliability of data acquisition and maximizing the potential of data, require attention. To solve these problems, a multimodal data-driven rehabilitation strategy auxiliary feedback method is proposed. In this study, depth sensor and functional near-infrared spectroscopy (fNIRS) were used to obtain ethology and brain function data, and skeleton tracking analysis and ethology discrete statistics were performed to assist the diagnostic feedback of rehabilitation strategies. This study takes rhythm rehabilitation training of autistic children as a case, and results show that the multimodal data-driven rehabilitation strategy auxiliary feedback method can provide effective feedback for individuals or groups. The proposed auxiliary decision method increases the dimension of data analysis and improves the reliability of analysis. Through discrete statistical results, the potential of data are maximized, thereby assisting the proposed rehabilitation strategy diagnostic feedback.