Virtual Reality Game-Based Adaptive Neurofeedback Training for Motor Imagery

Motor imagery-based brain-computer interfaces (MI-BCIs) are promising tools for motor rehabilitation applications. However, neurofeedback training (NFT) of traditional MI is boring and takes a long time. This study aimed to investigate whether adaptive NFT in virtual reality (VR) scenes could improve MI-BCI performance. In this study, we proposed a VR game-based MI-BCI paradigm that contained left and right upper limbs MI tasks. In the course of training, the system could adaptively adjust the game difficulty and dynamic output decoding result according to the training performance of individuals. Six subjects participated in the experiments. After three days of NFT, the activation of the motor cortex was enhanced. The average accuracy of MI tasks reached 83.52%, with an increase of 10.46%, compared with before training. The results demonstrate that the adaptive NFT system based on VR games proposed in this study has a good effect on MI performance and is expected to be used in rehabilitation training and treatment for stroke.