Research Studies:

STUDY 5: Computational Models of Stroke Neurorehabilitation: Mirror neurons, Observation and Learning

(Arbib and Schweighofer)

Study 5 will develop computational models of grasp and reach with the capacity to predict neuroplastic events related to cortical reorganization after stroke and following rehabilitation. The proposed modeling will focus on the control, acquisition, and re-acquisition after brain injury (such as strokes) of coordinated control programs for the visual and proprioceptive control of coordinated arm and hand movements (Jeannerod et al., 1995). This work will allow us a) to make specific and testable predictions on the organization and the reorganization after brain injury of the arm and hand motor circuits, and b) to propose new directions for the development of effective rehabilitation programs

Our strategy will be to emphasize initially a core model of reaching. This will be a basic model common to all species, strongly linked to data from our proposed rat studies as well as neural data from the primate literature. This will provide the base for a more elaborate study of the control of coordinated arm and hand movements in humans, culminating in the analysis of movement imitation in humans, in relation to rehabilitation strategies, in the later stages of this P20.


Part 1: Modeling interactions between cortex, basal ganglia (BG), and neuromodulators in acquisition, control and re-acquisition after brain injury, of coordinated arm and hand movements:

As stated above, the strategy will be to first develop the core multi-species model for reaching mechanisms then extend it to analysis of arm-hand coordination. Building on our previous work (see Fagg and Arbib 1998, Bischoff-Grethe, Crowley and Arbib, 2003), we aim at modeling a circuit for reaching and grasping that connects motor cortex, pre-SMA, premotor cortex, and the basal ganglia, all of which are implicated in reaching and grasping. We will then study the reorganization of this circuit after simulated lesions. Given the prevalent innervation of the basal ganglia and the motor cortical areas by dopaminergic neurons (Gaspar et al 1991), and previous work by McNeill (Hughs-Davis et al., 2004; Meshul et al., 2000) showing that dopamine is crucial in synaptic reorganization after brain injury, we will specifically address the roles of dopamine in the acquisition and re-acquisition of coordinated motor programs.


Part 2. Extension of the computational model from rats to humans to explore the role of observation in motor training - predictions on “rehabilitation via observation”.

The second goal of our study will be to develop models that make testable predictions useful to develop effective rehabilitation programs for humans. One attractive, yet untested, such rehabilitation program is “rehabilitation via observation”. For this purpose, we need to differentiate what is common to both rats and humans, and what is unique to humans. We must therefore offer hypotheses on what “lifts” the mechanisms that we study in rat to primates generally and humans in particular.

Our modeling of reaching and grasping in humans will bring in the mirror system to augment the basic rat model. Previous studies have shown that the ventral premotor cortex contains “mirror neurons” which are activated both during self-movements and during movement of others. Recently a tight link between the mirror neuron system and forward models has been suggested (Miall 2003). Forward models are neural circuits important for fast and accurate motor control such as reaching as they can predict the sensory consequence of motor commands. Thus, we will model the forward model for arm control, and incorporate this forward model in a fast feedback loop that will generate feedforward motor commands to the motor cortex (Miall et al., 1993). Further, according to our MNS model (Oztop and Arbib, 2002), the mirror neurons need to learn their response only when a positive grasp has been successfully performed, whether by the animal’s own movements and by the perception of other’s movements. Because dopaminergic neurons are known to carry reward signals, dopamine could be the gate to adapting the mirror neurons/forward models when a successful movement occurs (cf. Schweighofer et al., 2001, 2003). In a similar vein, norepinephrine (Schweighofer, Doya and Kuroda, 2004) could carry a gating error term between real and desired motor output that would have a role in “building a good map”, i.e., increasing the representation of the forward models in the area of workspace where high resolution is most needed (cf. Schweighofer et al. 2001).



Related Presentations and Publications by:
Dr. Arbib
Dr. Schweighofer


The Center is funded as part of the National Institutes of Health Roadmap Initiative, grant number P20 RR20700-01. NIH Program Administrator: Dr. Greg Faber