Phantom limb pain has been attributed to maladaptive plasticity after deafferentation. It has been hypothesized that reversing the maladaptive plasticity will relieve pain. Here, we tested the hypothesis that a restoration of upper limb functions using a brain-machine interface will normalize the maladapted cortical representation and relieve phantom limb pain.
A novel neuroprosthetic arm using real-time magnetroencephalographic (MEG) signals (Yokogawa) was applied to five patients with deafferentation pain due to brachial plexus root avulsion. First, in an open-loop session, they attempted to perform grasping or opening movements with their phantom hand every 5.5 s. The time-averaged MEG signals were used to train two decoders that predicted the performed movement type and the onset timing of the movement using support vector machine and Gaussian process regression. The prosthetic arm was controlled at the inferred movement onset to perform the predicted type of movement. Next, patients were instructed to control the prosthetic arm online for 10 minutes while watching the arm movement to improve their performance to control the arm. Finally, the open-loop session was performed again. The patient's pain was evaluated after each session by visual analog scale and the McGill Pain Questionnaire.
The estimated cortical currents on the contra-lesion sensorimotor cortex were significantly varied between the two types of movements (one-way ANOVA, p<0.05). The F-value of ANOVA was increased after the 10-minute training. Also, the classification accuracy using the contra-lesion cortical currents was significantly improved. Interestingly, the pain was exacerbated as the classification accuracies increased. These results demonstrate that, contrary to our hypothesis, restoring motor ability via a brain-machine interface can aggravate deafferentation pain.