Predicting the performance of motor imagery in stroke patients: multivariate pattern analysis of functional MRI data

In a brain-computer interface for stroke rehabilitation, motor imagery is a preferred means for providing a gateway to an effector action or behavior. However, stroke patients often exhibit failure to comply with motor imagery, and therefore their motor imagery performance is highly variable.

We sought to identify motor cortical areas responsible for motor imagery performance in stroke patients, specifically by using a multivariate pattern analysis of functional magnetic resonance imaging data.

We adopted an imaginary finger tapping task in which motor imagery performance could be monitored for 12 chronic stroke patients with subcortical infarcts and 12 age- and sex-matched healthy controls. We identified the typical activation pattern elicited for motor imagery in healthy controls, as computed over the voxels within each searchlight in the motor cortex. Then we measured the similarity of each individual's activation pattern to the typical activation pattern.

In terms of activation levels, the stroke patients showed no activation in the ipsilesional primary motor cortex (M1); in terms of activation patterns, they showed lower similarity to the typical activation pattern in the area than the healthy controls. Furthermore, the stroke patients were better able to perform motor imagery if their activation patterns in the bilateral supplementary motor areas and ipsilesional M1 were close to the typical activation pattern.

These findings suggest functional roles of the motor cortical areas for compliance with motor imagery in stroke, which can be applied to the implementation of motor imagery-based brain-computer interface for stroke rehabilitation.