1287_ch06.pdf

Conceptual Frameworks

for Interpreting Motor

Cortical Function: New

Insights from a Planar

Multiple-Joint Paradigm

Stephen H. Scott

CONTENTS

6.1 Introduction

6.2 Conceptual Framework

6.2.1 Sensorimotor Transformations

6.2.2 Internal Models

6.3 An Experimental Paradigm: Neural Correlates of Internal Models

6.4 Limb Mechanics

6.4.1 Neural Correlates of Limb Mechanics

6.5 Neural Correlates of Mechanic Loads

6.6 Summary and Conclusions

Acknowledgments

References

6.1 INTRODUCTION

In general, the problem solved by the brain is

how to convert visual information about the target location, initially sensed by

receptors in the retina, into motor action generated by temporal and spatial patterns

of muscle activities so as to stabilize the body and move the hand to the target. This

conversion of sensory to motor signals involves many cortical and subcortical regions

of the CNS, and a major focus of research is to identify the role played by each of

these regions.

0-8493-1287-6/05/$0.00+$1.50

Visually guided reaching is a natural motor task performed regularly by primates in

order to reach for and interact with objects of interest in the environment. The well-

deﬁned goal, moving the hand to a spatial location, makes it a popular paradigm

for exploring sensorimotor function.

to relate neural

activity to some features of behavior reﬂecting sensory, cognitive, or motor aspects

of the task.

How one chooses which variable to correlate depends highly on the conceptual

framework used to develop the experiment. This chapter starts with the important

issue of how theoretical concepts guide experimental design and data analysis.

Such

frameworks can be explicitly deﬁned, or in some cases, only implicitly imbedded

in the experiment and analysis. I will describe two conceptual frameworks for

interpreting neural activity during reaching: sensorimotor transformations and inter-

nal models. Both frameworks address the same biological problem: How does the

brain control the limb to reach toward a spatial target? The key difference is that

each framework focuses attention on a different aspect of the motor task and thus

each leads to different experiments. The sensorimotor transformations framework

has been used extensively over the past 20 years to guide neurophysiological exper-

iments on reaching, whereas the internal models framework has only recently had

an impact on experimental design.

The second half of this chapter illustrates how the notion of internal models can

be used to explore the neural basis of movement. A new experimental facility is

described that can sense and perturb multiple-joint planar movements and this is

followed by a brief description of the mechanics of limb movement. Finally some

preliminary observations are presented on neural correlates in the primary motor

cortex (M1) of the mechanical properties of the limb and of external mechanical

loads.

6.2 CONCEPTUAL FRAMEWORK

6.2.1 S

ENSORIMOTOR

RANSFORMATIONS

The

brain is assumed to convert visual information on target location into forelimb muscle

activation patterns through intermediary coordinate frames ﬁrst through various

kinematic representations of movement followed by kinetic representations. One

putative series of transformations is shown in Figure 6.1A , where spatial target

location is sequentially converted into hand kinematics, joint kinematics, joint mus-

cle torques, and, ﬁnally, muscle activation patterns. The use of intermediary repre-

sentations to plan and control movement seems like a reasonable assumption, par-

ticularly given the ubiquitous observation that hand trajectories are relatively straight

for point-to-point reaching movements.

3–5

6,7

Based on the concept of sensorimotor transformations, it seems obvious that the

key neurophysiological question is which coordinate frame is speciﬁed by the dis-

charge patterns of individual neurons in each brain region. Over the past 20 years

this framework has spawned myriad studies. As described below, some experiments

The basic question posed by studies that record neural activity during behavior

is this: What type of information is conveyed by the discharge pattern of individual

or populations of neurons? While cells are unlikely literally to code any engineering-

inspired variable, it is nonetheless valuable (and even necessary)

The most common framework for exploring the neural control of reaching has been

based on the idea of coordinate frames and sensorimotor transformations.

Target

Location

Hand

Kinematics

Joint

Kinematics

Joint

Torques

Muscle

Activity

Desired State

(Limb Position)

Internal Model

of Musculoskeletal

System

Muscle

Activity

Musculoskeletal

System

Limb Movement

Central

Nervous System

Sensory Feedback:

(Vision, Proprioception)

Two alternate frameworks for interpreting how the brain performs visually

guided movements. (A) The notion of sensorimotor transformations assumes that information

on spatial targets is converted into muscle activation patterns through a series of intermediary

representations. This framework leads to the scientiﬁc problem of identifying how these

representations are reﬂected in the discharge pattern of neurons in different brain regions.

(B) The idea of internal models is that neural processes mimic the properties of the muscu-

loskeletal system and physical objects in the environment. This framework leads to the

scientiﬁc problem of identifying how information related to the motor periphery and physical

loads is reﬂected in the discharge pattern of neurons.

have been designed to dissociate different variables, or levels of representation. In

other cases, a speciﬁc class of variable has been chosen

a priori

, either based on

the results of previous studies or simply for technical reasons.

One of the ﬁrst studies to record neural activity in the motor cortex during

reaching found that cell discharge was broadly tuned to the direction of hand motion.

This study showed that the cell discharge rate was maximal for movements in one

direction, the preferred direction (PD), and that the cell’s activity decreased as the

angle between movement direction and the cell’s PD increased. Further, the direction

of hand motion could be predicted from the discharge pattern of an ensemble of

neurons; this was termed the population vector hypothesis.

9,10

Variables of movement such as hand and joint motion are

highly intercorrelated, such that even if neural activity coded muscle velocity, one

would ﬁnd signiﬁcant correlations between cell discharge and hand motion. More-

over, the population vector will point in the direction of hand motion if three

conditions are met: (1) neural activity is symmetrically tuned to the direction of

movement; (2) the PDs of neurons are uniformly distributed; and (3) there is no

coupling between a cell’s PD and the magnitude of modulation during movement.

11–13

Any population of neurons that satisﬁes these conditions will predict the direction

of hand motion regardless of the underlying information conveyed in its discharge

FIGURE 6.1

A criticism often levied upon these studies has been that these hand-based

correlates could be observed regardless of the type of information conveyed by

individual neurons.

reiterated this point by illustrating

how a broad range of observations between hand movement and neural activity, both

at the single-cell and at the population level, could be explained if cells were simply

coding multidimensional muscle activation patterns. While the correct explanation

of the precise details of all hand-based correlations is a matter of debate,

the

article by Todorov illustrates how difﬁcult it is to interpret the discharge of neurons

with simple correlation methods.

In spite of these concerns, a school of thought was created around the population

vector hypothesis and the notion that neural activity in M1 during reaching should

be interpreted using hand-based variables. Studies illustrated that neural activity in

M1 and other sensorimotor areas correlates with the direction of hand motion, hand

velocity, movement extent, and end position.

14–17

These studies illustrate that neural

activity certainly correlates with these hand-based or end-point variables, but in

many cases such activity may actually reﬂect relationships to the sensory and motor

periphery such as motor patterns at a single joint or multiple joints.

18–21

11,13

Few believe neural activity in M1 is coding the activity of single muscles, but

the hand-based framework makes a substantive leap away from the motor periphery.

In the extreme, descending commands are assumed to convey only the direction of

hand movement that gets converted into motor output at the spinal level.

23,24

although such inter-

pretations remain controversial.

25–27

A key feature of many studies has been to dissociate explicitly different features

of the task, such as sensory versus motor,

28,29

global variables (hand–target) versus

For example, we per-

formed an experiment where reaching movements were performed to the same target

locations but with two different arm postures: ﬁrst, in a natural arm posture where

the elbow tended to remain directly below the hand and shoulder, and second, in an

abducted posture with the shoulder abducted and with the elbow almost at the level

of the hand and shoulder.

30–36

or kinematic versus kinetic variables.

37–39

This task dissociated global features of the task related

to the spatial target and hand motion, which remained constant, from joint-based

variables related to joint motion, joint torque, or muscle activity, which varied

between arm orientations. We found that most neurons showed changes in activity

either by changing their directional tuning or by modulating the overall level of

activity, suggesting that neural discharge was related in some way to the motor

periphery. Some cells, however, showed no changes in activity when movements

were performed with the two arm orientations. Such invariances could reﬂect that these

cells are specifying global features of the task, although it is still possible that such

cells could reﬂect joint-based information. (See Scott and Kalaska.

)

A cleaner dissociation between muscle- and hand-based features of movement

was provided in a study where wrist movements were performed with three different

forearm orientations: neutral, supinated, and pronated .

They found that some cells

varied their directional tuning in a manner that was similar to the variation observed

for muscles, whereas others showed no change in directional tuning, as would be

expected if neurons reﬂected the spatial direction of the task. However, most of these

patterns. A recent theoretical study by Todorov

The shift

away from the motor periphery has been extended further to suggest that M1 may

be involved in cognitive processing such as mental rotation

joint–muscle,

The obvious conclusion is that there appears to be no

single uniﬁed coordinate frame in M1. This of course causes considerable problems

for the population vector hypothesis, which presupposes that a global signal related

to the direction of hand motion is created across the cell population. In M1, cells

respond to many different variables, with some cells largely reﬂecting kinematic

features of the task and other neurons reﬂecting kinetic features. Whenever force

but not kinematic motion is modiﬁed, these latter cells, which modulate their activity

with force output, will alter estimates of hand motion.

While most agree that neural activity in M1 reﬂects a mixture of different

kinematic and kinetic features of movement, the notion that the brain performs a

series of sensorimotor transformations to execute reaching movements assumes a

certain relationship between these representations. Speciﬁcally, cells insensitive to

force output are assumed to reﬂect a higher level representation of movement which

gets converted by cortical processing into a lower level representation; cells sensitive

to force output are classiﬁed as this lower level representation. Are cells that are

insensitive to force output necessarily reﬂecting a higher level representation than

cells that are sensitive to force output? This assumption would seem reasonable, if

muscle activity (electromylography [EMG]) were the only feature of motor behavior

controlled by the brain.

However, descending commands to the spinal cord must consider more than just

muscle activity.

1,20,40

Alpha motoneurons, which innervate extrafusal muscle ﬁbers

and produce force, represent only one type of motoneuron. In each motoneuron pool,

there is a large number of gamma motoneurons that innervate intrafusal ﬁbers in muscle

spindles,

1,41,42

which may be equal in proportion to alpha motoneurons in some muscles.

There are even beta motoneurons innervating both intra- and extrafusal muscle

ﬁbers.

Another role for descending commands is to modulate and inﬂuence sensory

feedback.

Spinal reﬂexes can also create various contingency plans for unexpected

perturbations or errors,

which must also be selected or modiﬁed by descending

commands. It is quite possible that up to two thirds of descending signals from the

cortex to the spinal cord are related to controlling these other features of motor output.

However, little is known about cortical discharge related to controlling gamma-

motoneuron activity and spinal reﬂexes during volitional tasks since experimental

paradigms, including our own, tend to focus on alpha-motoneuron activity.

It is

quite possible that neurons related to these other features are relatively insensitive

to variations in force output during motor behavior. Within the rubric of sensorimotor

transformations, such neurons would be assumed to code a higher level representa-

tion of movement related to the kinematic features of the task when in fact they

were simply involved in controlling relatively low-level but non-EMG features of

the task. Furthermore, such discrete segregation between alpha-motoneuron activity

and other spinal processing is highly unlikely and descending signals likely reﬂect

a mixture of inﬂuences on spinal circuitry.

47,48

latter spatial/hand cells still showed changes in the magnitude of activity for move-

ments with different forearm orientations.

All these studies illustrate that primary motor cortical activity correlates to

almost every imaginable task variable, including spatial target location, hand move-

ment direction and extent, hand velocity, joint velocity, force output, and muscle

activity, to name a few.

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: