Real-time classification and feedback of motor intention using motor cortical activity patterns during motor imagery

Abstract

Recent studies have shown that motor imagery training improves the actual performance of motor skills, which are involved in an intentional movement to perform a specific task. The effect of training is related to the plasticity of neural activity of motor intentions. However, it is still unknown how we practice motor imagery in order to make our desired activity patterns. Here, we first report real-time fMRI classification system can classify brain states related to motor intentions during motor imagery and give feedback about imagery to subjects. To generate accurate feedback, we propose a novel detrending for improving classification performance. Motor imagery training also has showed the stability of neural activity patterns related to motor intentions during motor imagery. Motor execution and motor imagery classification using simulated MVPA revealed the representation of motor information as well as the feasibility of real-time classification. After feedback training, increased motor information and motor network showed that real-time classification of fMRI signals and feedback could be helpful to self-regulation of activity patterns for consistency. Unlike previous studies, we could classify brain states during complex movements imagery. These results indicate that our system could be a useful tool in certain applications including rehabilitation and fMRI-BCI for the improvement of motor skills as well as function.