Elsevier

Neural Networks

Volume 48, December 2013, Pages 25-31
Neural Networks

A model of task-specific focal dystonia

https://doi.org/10.1016/j.neunet.2013.06.012Get rights and content

Abstract

Task-specific focal dystonia is a task-specific movement disorder which manifests itself as a loss of voluntary motor control in extensively trained movements. The condition is most frequent in musicians. Until today, the aetiology of focal hand dystonia is not completely understood, but there is growing evidence for an abnormal cortical processing of sensory information, as well as degraded representation of motor functions. It was demonstrated that in the somatosensory cortex the topographical location of sensory inputs from individual fingers is corrupted. Occasionally, a change in sensory information of the hand may at least temporarily improve the condition. This phenomenon is called sensory trick. In this paper, we propose a model of encoding of sensory stimuli which could explain the task specificity of cortical representations of the fingers or other effectors in the context of dystonia. In the framework of this model a sensory stimulus is encoded as a signal vector of higher dimension. A part of its components directly represents the sensory stimulus, while the remaining components describe the context. This model does not only account for the task specificity, but may also explain some characteristics of the retraining process in this disorder.

Introduction

The general term dystonia is used to describe a syndrome characterized by involuntary sustained muscle contractions, frequently causing twisting and repetitive movements, or abnormal postures (Jankovic, 2005). If these symptoms are restricted to one body part, the syndrome is termed “focal dystonia”. In task-specific focal dystonia the most prominent characteristic is the degradation and loss of voluntary control of highly overlearned complex and skilled movement patterns in a specific sensory-motor task. The highest incidence of focal task related dystonia is found in musicians (Altenmüller & Jabusch, 2009). For example, in pianists suffering from hand dystonia, piano playing might be severely compromised due to loss of fine motor control of individual finger movements; however, writing on a computer keyboard in high speed using ten finger system is not deteriorated. Interestingly, musician’s dystonia seems to be clearly related to intense practice of the disturbed movement patterns. For example, fast, repetitive complex movements of the left hand in violinists are more often affected as compared to the more balancing movements of the right bowing hand. Vice versa, in pianists, dystonic movements are predominantly developed in the right hand, which is at least in the baroque, classical and romantic period more challenged as compared to the left due to the more frequent occurrence of fast figurations, trills and scales (Jabusch, Zschucke, Schmidt, Schuele, & Altenmüller, 2005). Although structural abnormalities (Black, Ongur, & Perlmutter, 1998) as well as hereditary factors (Schmidt et al., 2009) have been recently identified, it is generally agreed that overtraining plays an important role in this type of dystonia (Baur, Jabusch, & Altenmüller, 2011). This has impressively been demonstrated in an animal experiment with monkeys trained to do extensive fine motor movements of the hand developing a dystonia-like syndrome (Byl, Merzenich, & Jenkins, 1996). Consequently, it had been suggested that focal dystonia has to do with a misled self organization process. Sanger and Merzenich for example have hypothesized that it is a manifestation of an unstable sensorimotor control loop. They discussed several mechanisms which possibly could lead to a gain >1 in the feedback loop and they formulated a computational model which explains some aspects of focal dystonia (Sanger & Merzenich, 2000). For most of the characteristic phenomena, however, a satisfying theoretical explanation does not yet exist. Amongst them are the following.

  • In some musicians with focal dystonia it could be observed that the somatotopic order of the somatosensory finger representations in the cortex was distorted and that the receptive fields of single fingers overlapped (Elbert et al., 1998). This may suggest that the origin is not so much instability of the control loop but primarily a corrupted mapping of sensory signals to cortical regions. In writer’s cramp such a distorted mapping has also been demonstrated in the Basal Ganglia, more specifically in the Putamen (Delmaire et al., 2005). It has been hypothesized that such a distortion of somatotopic maps is due to maladaptive plasticity and inefficient lateral inhibition, leading to a fusion of previously separated sensory-motor receptive fields. (For a review see Hallett, 1998.)

  • Musician’s dystonia is in most instances task specific. As mentioned above, the condition affects the control of finger movements predominantly in the context of instrument playing. Astonishingly, it does not impair the function of the same fingers in other activities, although at closer examination minor abnormalities not related to a specific task can become obvious (Byl et al., 1996, Rosset-Llobet et al., 2007). This means that generally the corrupted mapping is effective only in a special context.

Among the neural networks studied in computer science there is a type which lends itself for modelling cortical mappings and particularly the self organization processes which generate them: the Kohonen network (self organized mapping, SOM, Kohonen, 1982, Kohonen, 1995) is an array of ‘theoretical’ neurons which processes a set X={xi} of training signals in such a way that to each signal xi of the set X a neuron is assigned, i.e. the signal set X is mapped to the array of neurons. An important property of this mapping is that similar signals x,x are mapped to neighbouring neurons. In terms of mathematics this property is called preservation of neighbourhood. It closely corresponds to somatotopy in neural organization.

Soon after the first SOM publications (Merzenich et al., 1984) have shown that the reorganization observed in the sensory cortex of mammals after peripheral nerve damage can be modelled as an adaptation process in a Kohonen map. Martinetz, Ritter, and Schulten (1988) demonstrated that the auditory cortex of a bat may be considered as a neighbourhood preserving map of the relevant space of ultrasonic signals. Meanwhile the SOM concept is accepted as a tool for understanding the processing of sensory signals in the brain (Sirosh and Miikkulainen, 1995, Turrigiano and Nelson, 2004, Wiemer et al., 2000).

The fact that after amputation of a digit the corresponding cortical region by and by takes over new functions was impressingly modelled in the simulation experiment (Merzenich et al., 1984). Yet this raises the question how stimuli are encoded as signals and whether the mapping of different stimuli σ,σ to different neurons N,N implies a corresponding wiring of signal paths. If there were simple signal paths from sensors to neurons of the cortex then it were difficult to explain how the connections are modified during the self organization process and how they are switched to another cortical region after an amputation. A more plausible assumption is that each stimulus from a certain body region is encoded as a set of simultaneous signals which are transmitted to the cortex via one and the same bundle of paths. A similar assumption was already made in the paper by Merzenich et al. (1984). Here a stimulus is represented by a two-component signal vector x=(x1,x2) where x1,x2 are the coordinates describing the location of the stimulated sensor. This assumption was justified within the scope of a principal demonstration. However the (Cartesian) coordinate pair x1,x2 refers to a pair of rectangular axes which have no natural relationship to biological structures. It is not plausible that a sensor in the hand surface will produce signals x1,x2 which depend on the choice of a special coordinate system.

Therefore we propose a simple model which explains the encoding of stimuli as high-dimensional signal vectors x. This is the reason why a sensory stimulus was denoted above not as a single signal x but as a signal vector x=(x1,x2,,xk). The encoding of stimuli is based solely on the concept of distance. It does not refer to a special coordinate system.

If one accepts the hypothesis that each stimulus is represented by a high-dimensional vector x then it needs only a slight generalization of this picture to explain the task specificity of focal dystonias.

The function principle of self organized mappings is sketched in Section  2. Here naturally we must restrict ourselves to those features which are important in the context of the present paper. Section  3 describes the model of stimulus encoding, i.e. the generation of a signal vector x for a given stimulus σ. The model is tested by a simulation experiment which in a certain sense is a remake of the experiment by Merzenich et al. (1984). It reveals that the neighbourhood preserving mapping between hand and sensory cortex as well as the reorganization of this mapping after amputation of a digit is correctly reproduced. In Section  4 it is shown that application of the same principles to more general types of stimuli leads to an explanation of task specificity. This is not only of theoretical interest. In attempts to cure musician’s dystonia by retraining, one of the problems is the so called deprogramming, i.e. the destruction of a corrupted cortical representation. If the mechanism of task specificity were known this would facilitate the planning of deprogramming strategies.

Section snippets

Cortical representations as self organized mappings: Kohonen networks

A structure which lends itself for modelling cortical representations is the Kohonen network (Self Organized Mapping, SOM). It is a mapping of a signal set XRm to a regular, typically two-dimensional grid of n×n simulated neurons Np,q. A signal xX is given in parallel to all neurons and the neuron Np,q with maximal output y, the so called winner neuron, is considered the picture of x. We will denote it by Npc,qc. The mapping is neighbourhood preserving and since preservation of

What is transmitted to the sensory cortex? Encoding of stimuli as signal vectors

Merzenich et al. (1984) have studied the cortical representation of the hand in monkeys. In a simulation they demonstrated that the reorganization process after amputation of a finger can be modelled by a Kohonen SOM. In particular they could reproduce the experimental findings that the area belonging to the amputated finger is gradually occupied by the representations of the remaining fingers.

We performed a similar simulation experiment for two reasons.

  • 1.

    Since we conjecture that the task

Task specificity of focal dystonia

The model proposed in the previous section suggests a generalization: the m-component signal vector x comprises sensory signals as well as signals from other origins. Since the order of components is irrelevant one may consider the input to the cortex as a vector x=(x1,,xm)=(s1,,sk,c1,,cmk)=(s,c) where s1,,sk are sensory signals. Let us assume for the moment that c1,,cmk are signals describing different contexts like writing (W) or violin playing (V). Then c may take two different

Strategies for deprogramming and retraining

Musician’s dystonia and focal dystonia in general are difficult to treat. Injections of Botulinum Toxin into the cramping muscles and anticholinergic drugs may produce temporary relief and improvement of symptoms (Schuele, Jabusch, Lederman, & Altenmüller, 2005), but these methods do not provide a definite cure of the underlying causal dysfunctional mapping (Altenmüller & Jabusch, 2008). In task-specific hand dystonia, there have been several retraining strategies developed, attempting to

A possible cause of focal dystonia

Little is known about the physiological conditions or processes causing a focal dystonia. Musicians are a group of individuals which regularly performs tasks with extreme velocity requirements. As mentioned in the introduction, limbs executing fast repetitive movements are more frequently affected than those executing rather balanced tasks (Altenmüller, Baur, Hofmann, Lim, & Jabusch, 2012). This suggests that focal dystonia has to do with velocity and with high signalling rates. Each neuron has

References (32)

  • N.N. Byl et al.

    Sensory dysfunction associated with repetitive strain injuries of tendinitis and focal hand dystonia: a comparative study

    Journal of Orthopaedic and Sports Physical Therapy

    (1996)
  • V. Candia et al.

    Effective behavioural treatment of focal hand dystonia in musicians alters somatosensory cortical organization

    Proceedings of the National Academy of Sciences of the United States of America

    (2003)
  • C. Delmaire et al.

    Disorganized somatotopy in the putamen of patients with focal hand dystonia

    Neurology

    (2005)
  • T. Elbert et al.

    Alteration of digital representations in somatosensory cortex in focal hand dystonia

    NeuroReport

    (1998)
  • H. Flor et al.

    Phantom-limb pain as a perceptual correlates of cortical reorganization following arm amputation

    Nature

    (1995)
  • M. Hallett

    Physiology of Dystonia

    Advances in Neurology

    (1998)
  • Cited by (11)

    • Experience-dependent neuroplasticity in trained musicians modulates the effects of chronic pain on insula-based networks – A resting-state fMRI study

      2019, NeuroImage
      Citation Excerpt :

      Particularly, as extensive practice routines can also lead to the development of focal dystonia (Altenmüller, 2003) and chronic pain (Steinmetz et al., 2015). While compromised motor function in focal dystonia has been attributed to maladaptive changes in the brain (Altenmüller et al., 2015; Altenmüller and Muller, 2013), however, the pathophysiology underlying chronic pain in skilled musicians remains unknown. Epidemiological studies indicate that about 80% of musicians experience musculoskeletal pain syndromes throughout their careers (Cruder et al., 2018; Kok et al., 2016; Steinmetz et al., 2015), an incidence rate that is about 60% higher than in the general population (Breivik et al., 2006) and that it is even higher in music students (Brandfonbrener, 2009; Steinmetz et al., 2012).

    • Brain Plasticity in fMRI and DTI

      2022, Medical Radiology
    • Treatment of musicians’ dystonia

      2019, Current Clinical Neurology
    • Expertise in Music

      2018, The Cambridge Handbook of Expertise and Expert Performance: Second Edition
    View all citing articles on Scopus
    View full text