Stroke is a leading cause of neurological disability among adults and often leads to functional deficits in motor control. It is then known that motor intention and sensory feedbacks are an effective method of improving damaged motor functions, especially when both are synchronized. There are several tools to detect motor intention, such as EEG, EMG, NIRS etc which are developed by brain computer interface (BCI) systems. Sensory feedbacks can be given by functional electrical stimulation (FES), haptic interfaces, robotic devices. This talk will first give an overview of them. Then I will present an EEG/ERD modulated FES system that has better temporal synchronicity and correlation between motor intention and proprioceptive feedbacks in comparison to the previous systems. When this system was applied to the affected lower limb of a paretic stroke patient, we observed short-term functional improvement after the training using ERD modulated FES system.

Neural plasticity induced by rehabilitative training plays an important role in functional recovery of paretic limb in stroke patients. The region of the brain responsible for recovery is expected to differ depending on the locus and size of the lesion. Since there have been no way to monitor brain activity during rehabilitative training of paretic limb, the way how activity change during the course of recovery is not yet well known. I will introduce our recent challenge to estimate brain activity during hand rehabilitation by combining near infrared spectroscopy (NIRS) and electroencephalogram (EEG).

One of the main problems for rehabilitation is to extract the intention of patients. For example, after stroke the patient may have affected the arm movement. Then, he/she can not move their arm as he wants. To support his movement, the timing, direction, amplitude, speed and so on, are needed. Muscle activities are one candidate for estimate these properties. But it is difficult to measure them from the stroke patients. In this case, EEG is another signal to extract the intention. ERD/ERS is used for extraction of the intention. Timing may be extract by it, but the direction, speed, amplitude would be difficult. We are trying to reconstruct EMG activities from EEG signals using hierarchical base method for estimation of neural current source signals. Also the device which is controlled by EMG signals for controlling the equilibrium point and stiffness, has been developing. These two techniques will be combined for the rehabilitation robot.

Functional tasks performed by humans are stereotyped and use only a few degrees-of-freedom (DOF). In order to avoid complex robotic trainers, it has been proposed to use devices involving only these DOF. To overcome limitations in learning due to the resulting mechanical constraints, we provide visual feedback of the movement corresponding to the force measured along the constraints. This talk will first present results of healthy subjects learning a curl force field while performing arm movements in a channel. The subjects are able to learn from virtual visual error, despite zero proprioceptive error and conflicting sensory signals. We will then present initial results of using this strategy to learn movements of stroke patients moving in a channel. Finally, we will examine virtual learning using a force sensor, which is used to produce virtual movement without actual movement.

Animal models of stroke and correlated human studies demonstrate that functional recovery of motor patterns after stroke is obtained through the use-dependent reorganization of neural mechanisms, exploiting basic properties of neural plasticity. However, it is not movement per se, obtained for example by means of passive mobilization, which is effective in recruiting plastic adaptation. The key is movement associated with a task and a volitional effort. In fact, task oriented training has emerged as a leading concept in clinical practice. The question then is: How can volitional effort be evaluated? How can it be enhanced? These issues are discussed with reference to three important implementation aspects: 1) level of stiffness of robot assistance; 2) modulation of the patient's stiffness during functional recovery; 3) enhancing proprioceptive awareness.