Goodwin Lab ~ Hand Function

Overview

The evolutionary development of the hand, and the associated neural sensorimotor system, has been central in the success of primates and of humans in particular. The hand is part of a sophisticated feedback control system that allows us to grasp, explore and manipulate objects in our environment with great dexterity and precision.

There are many diseases, ranging from peripheral neuropathies to stroke, that affect hand function. In patients with such diseases, control of hand movements is severely compromised and they are incapable of the precise manipulations that are critical in our everyday life. Yet in spite of the obvious importance of hand function, and despite the biological elegance of this sensorimotor device, very little is understood about the neural mechanisms that are at the core of the system.

top of page

Approach

Our interest is in understanding the neural control mechanisms responsible for precision manipulations particularly in regard to the tactile sensory information relayed from the hand to the brain. In order to do this, we use a combination of techniques - recording from single nerve cells at various levels of the primate nervous system; measuring human tactile capacities using psychophysical methods; and finally we integrate these two components by constructing neural models of the system.

The depth of anaesthesia of our experimental animals together with their vital signs are monitored continuously using equipment similar to that used during surgery on humans.

Tactile mechanisms underlying grasp stability

Some of the remarkable features of the sensorimotor system become apparent when considering the simple task of lifting a glass of liquid vertically off a table. The weight of the glass imposes a load force on the digits which acts in a direction tangential to the skin. Voluntary contraction of muscles controlling the hand produces a grip force which acts in a direction orthogonal to the skin. These two forces, which are at right angles to each other, are completely different. The load force is not within our control; it depends on the weight of the object which, in general, we do not know. It is also subject to unpredictable disturbances. In contrast, the grip force is entirely within our control and is set by efferent signals from the central nervous system. Ultimately, we have to use grip forces that are appropriate moment to moment for the task. The crux of the problem is that the required grip force depends on the weight of the glass and on its effective coefficient of friction with the skin, neither of which are known to the person lifting it.

top of page

Neural Population Responses

Force coding

We have measured human ability to scale these two force components and we are currently addressing the issue of how tangential 'load' forces and torques are encoded by the cutaneous afferent population. Our hypothesis was that the SAI afferent population would encode the magnitudes and directions of the linear load forces. We find that in the monkey, SAI afferents respond vigorously to tangential load forces in a direction-selective manner (as shown in the polar plots). Quantitative population reconstructions have shown that the SAI population is able to extract the magnitude and direction of load forces independently of grip forces.

Responses of two SAI afferents innervating different locations on the fingerpad to forces applied sequentially in 8 directions.

Object feature coding

Whether manipulating objects or exploring our environment, we also require detailed information about the objects we are contacting. Information about essential characteristics such as shape is encoded in the multidimensional response elicited by contact with the object in the population of mechanoreceptive afferent fibres innervating the skin. The first step in defining these population responses is to record the responses of single fibres in the peripheral nerves of anaesthetised monkeys and then, from multiple single fibre responses, reconstruct the population response. Changes in object shape are reflected in variation in the shape of the response profile across the population of afferents activated by contact with the object. These responses can be characterised mathematically and we do this in order to examine more complex aspects of population responses such as variance and differences in sensitivity.

Neural population reconstructions of responses to curved objects (left) and a surface with different gap widths.

top of page

Human Performance

In order to interpret the neural responses we measure, it is necessary to determine human capacities precisely. We use standard psychophysical methods to obtain accurate measurements of human performance: magnitude estimation for scaling the perceived magnitude of a stimulus variable over a given range, and signal detection theory analysis of two-alternative forced-choice paradigms, which yield bias-free estimates of the acuity of the tactile system. The range of human performance we study includes those capacities which underlie dexterous hand function.

Shape & Orientation

By way of illustration consider the seemingly simple task - writing. When using a pen, its orientation within the grasp as well as its shape and contact force must be signalled to the sensorimotor control system in order to generate the intricate movements needed for writing. Psychophysical measurements have shown that humans can discriminate differences of around 5 degrees in the orientation of cylindrical objects. Acuity is greater for less curved (larger radius) objects; (this threshold is indicated by the dotted lines - discrimination index value of 1.35).

Human ability to discriminate the orientation of a cylindrical object.

Such things as the object's shape and its position and orientation on the skin are critical parameters determining precise manipulation. The sensorimotor system also needs to coordinate the three-dimensional forces which act at the finger/object interface in order to optimize and maintain grasp stability.

Computer-controlled stimulator capable of delivering a range of forces and recording reactive forces both perpendicular and tangential to the surface of the fingerpad.

Grip & load force

In order to ensure grasp stability, the sensorimotor systems requires detailed information about the magnitude and direction of load and rotational forces (tangential to skin surface) plus feedback about grip force (normal to skin). Previous studies have shown there is rapid, precise coordination between grip and load forces which deteriorates with digital nerve block. Using a paradigm that simulated natural manipulations, we quantified human capacity to scale the magnitude of tangential and normal forces and torques using only cutaneous information. Our subject could scale these different force components independently.

These data show that the cutaneous afferents are able to provide a wealth of precise and independent information about object and task characteristics. Our neural experiments are providing data which suggest possible coding mechanisms to explain these findings.

top of page

Neural Models

Mathematical simulation

The mathematical modelling approach we use in this final integrative phase allows us to address a number of questions which relate to the signalling capacity of whole populations of afferent fibres acting in concert. By simulating the responses of populations of afferent fibres activated by a stimulus with a number of varying parameters, we can examine how variability in intrinsic population characteristics impacts on the acuity of the tactile system. This is an essential step if we are to understand the extent to which, and in deed how, the tactile system is able to extract precise and detailed information about an unknown stimulus from a multi-dimensional neural signal.

The mathematical model we have constructed emulates real neural populations in terms of the distribution of fibre sensitivities within and between populations, innervation densities and geometric configuration, and response variability and covariance. The 3D construction illustrated here represents the response profile of a realistic afferent population when a spherical stimulus (curvature 256/m) was presented at a prescribed location on a hypothetical finger. This is to be compared with that for an ideal population with uniform sensitivity and no response noise (inset). In spite of the distortions in the shape of the more realistic response profile, changes in stimulus shape can be extracted from the population response with a level of acuity which correlates well with human performance. The model achieves this by using a relatively simple algorithm as a 'neural measure' which describes the shape of the whole population response.

Neural networks

It is becoming increasingly obvious that neurons in the brain are highly interconnected, and that much of the computational power of the brain is achieved by the combined activity of these interconnected neurons. Information may be represented in the activity of such networks although not readily seen in the responses of single neurons. In order to gain insight into how such processes may work, we are using neural network models to relate the response properties of primary afferent fibres (the inputs to the network) and the capacities of human subjects (the output of the network). The figure depicted here shows a simplified diagram describing a back-propagation neural network. By selectively 'knocking out' certain hidden layer units, we can evaluate how output functions are affected. We find that some hidden layer units are critical for shape discrimination whilst others are critical for localising the object.

Schema of a neural network determining object shape and position.