Feasibility Study

Coordination of Direct/Inverse Approaches in Brain

Kenji Kawano

■Object

In the study of sophisticated information processing functions of the brain, some important findings have been obtained in recent years with regard to vision, motor control and long-term memory. However, they are still fragmentary, phenomenological, and lacking in common theory to integrate them. In the present F/S, the brain research will be deployed on the basis of an entirely new concept by regarding the brain as a ﾒinverse problemﾓ solving device (Fig. 1). This is an innovative hypothesis, proposed and being underway toward demonstration by our research group for the first time in the world. The information processing in the brain proceeds through interaction with the outer world. The ﾒdirect problemﾓ is basically given from the outside (Fig. 1). Since most of information is usually lost when the problem is fed into the brain, the brain is forced to solve the ﾒdirect (forward) problemﾓ in reversed direction based on insufficient information. If the brain is regarded as an ﾒinverse (backward) problemﾓ solving device, a clue for a scientific clarification of not only sensory and motor functions but also higher brain functions such as thought and emotion should become available. In the present F/S, we will attempt to define the basic principle of the information processing in the brain, which is an ﾒinverse problemﾓ solver to solve inverse problem autonomously in coordination with the "direct problem."

Fig. 1 Schematic diagram of the coordination of inverse/direct problems in the brain

■Results in Fiscal Year

The main results obtained in this F/S can be related to two themes: 1) Feed-forward control in eye movements; and 2) adaptive mechanisms in reaching movements.

1) Feed-forward control in eye movements

Movements of a visual scene evoke short-latency ocular-following responses. In our laboratory, we have focused on understanding the neural mediation of ocular following. Evidence from single-unit recordings and focal chemical lesions has suggested that early ocular following is mediated by a pathway that includes the medial superior temporal (MST) area of the cortex, the dorsolateral pontine nucleus (DLPN), and the ventral parafloccular (VPFL) lobes of the cerebellum. To characterize the response properties of the simple spike activity of the P-cells in the VPFL, we reconstructed their temporal firing patterns based on the inverse-dynamics representation of the simultaneously recorded eye movements. The result that simple linear inverse-dynamics representation can satisfactorily predict complex time courses of the Purkinje cell firing patterns suggests that the simple spike firing frequency of these Purkinje cells encodes the motor commands for the dynamic component (eye velocity and acceleration) of the ocular-following responses. On the other hand, when the same analysis was applied to the responses of the neurons in the MST and DLPN, most of the neurons in these regions were not predicted by a simple linear inverse-dynamics representation of the eye movements. The results are consistent with the idea that the cerebellum may be a major site of the inverse dynamics model for controlled movements.

To understand how the inverse dynamics model is constructed in the cerebellum, we studied the complex spike activity of the Purkinje cells in the VPFL during ocular following. When averaged over many trials, the discharge-frequency profile of the complex spikes is almost a mirror image of the simple spike profile and can also be well modeled by the inverse-dynamics representation of the eye movements. This indicates that the information encoded in these averaged complex spike discharges matches the spatial and temporal characteristics of the information encoded in the simple spike discharges exactly as required to construct the inverse dynamics model in the cerebellum.

2) Adaptive mechanisms in reaching movements

Reaching toward a visual target accurately is disturbed after the visual field is displaced by prisms, but recovers with practice. When the prisms are removed, human subjects misreach in the opposite direction (after-effect). We examined the relation between the size of the after-effect and the movement velocities, both during and after the visual displacement. Trained subjects were required to reach with one of 4 movement durations (250 - 5000 msec) from a fixed starting point to a target that appeared at a random location on a tangent screen. The size of the after-effect was the largest when the movement velocities during and after the visual displacement were the same. It became smaller as the difference in velocity became larger. When the contralateral arm was used after visual displacement, the after-effect was never significant. Because the adaptation is selective to velocities and to each other arm, we infer that the underlying changes occur at a late stage in the transformation from visual input to motor output.。