All About Parkinson's Disease

All About Parkinson's Disease By Lianna Marie

Get Instant Access


An interesting aspect of Basal Ganglia (BG) function is how activity in the widespread cortical projections to the striatum (numbering ~100 times the number of striatal neurons (Oorschot, 1996)) can be efficiently summarized (compressed). Computational neural net models that perform this type of statistical operation, called dimension reduction, attempt to remove correlations from output neuronal units (Foldiak, 1990). Uncor-related outputs result in the most efficient means for information transfer (Nadal and Parga, 1994), as highly correlated outputs implies redundancy, and therefore inefficient compression (Bell and Sejnowski, 1995). Some computational models of the striatum and basal ganglia incorporate more than simple dimension reduction; they include a phasic dopaminergic component which adds behaviour saliency to the process (Bar-Gad et al., 2000; Bar-Gad and Bergman, 2001).

Consistent with theoretical results, experimental evidence confirms that primate BG neurons normally tend to fire independently from one another during motor tasks (Bar-Gad et al., 2003). Correlations between neighbouring neurons in the primate globus pallidus are usually very weak, however, after MPTP application, neighbouring neurons in GPi dramatically increase their correlation, and to a lesser extent neighbouring neurons in GPe. The ineffective dimension reduction in the Parkinsonian state may result in abnormal feedback within the basal ganglia loops and contribute to widespread pathological oscillations throughout the basal ganglia (Bar-Gad and Bergman, 2001; Raz et al., 2001; Bar-Gad et al., 2003).

Abnormal synchronization within the basal ganglia may, in fact, be a characteristic of the Parkinsonian state. Raz et al. (2001) examined the relation between tonically-active neurons in the striatum and pallidum in the vervet monkey, both before and after injection with MPTP. After MPTP injection, there was a marked increase in the number of neuron pairs which displayed significant peaks in cross-correlograms corresponding to ~10Hz. They concluded that ''coherent oscillations of the whole basal ganglia circuitry underlie the clinical features of Parkinson's disease'' (Raz et al., 2001).

A cardinal symptom of PD that may be a consequence of pathological BG oscillations is bradykinesia (Raz et al., 2001). However, the clinical sign of bradykinesia is merely an indicator for much broader areas of motor disability in PD such as gait disturbances and micrographia. Since pathological BG oscillations have been suggested to be the key underlying feature of many of the clinical manifestations of the disease (Raz et al.,

2001), demonstration of a quantitative measure indicative of these oscillations (e.g. abnormal increase in ~10Hz EEG/EMG coupling) that can be obtained non-invasively and inexpensively would be beneficial. Demonstration of an electrophysiological marker for bradykinesia will provide a target to monitor various pharmacological and surgical therapies.

Oral pharmacotherapy is typically used in treatment in PD, and while this may partly reverse tonic dopamine levels (Heimer et al.,

2002), it will not be able to provide the precise phasic changes normally a feature of dopaminergic neuronal firing. Yet in other dopaminergic systems, such as the ventral tegmental/prefrontal region, dopaminergic neurons demonstrate precise timing of their firing patterns, especially with respect to expectation of reward (Schultz et al., 1997). In contrast, Deep Brain Stimulation (DBS) methods would theoretically have the capability to modulate their stimulation on a second-by-second basis if the appropriate cues for doing so could be determined.

A non-invasive assay of abnormal BG oscillations influencing motor cortex in PD patients would hence prove valuable. It would enable an assessment of how much of the motor deficits commonly observed in PD are due to pathological BG oscillations that are observed in animal models of PD, and the influences of tonic drug therapy. Moreover, because measurements of oscillations could be done on a second-by-second basis, behavioural paradigms could be developed to determine the properties of different sensory stimuli that may result in abnormal BG oscillations (or a pathological increase in normally-present physiological oscillations).

Corticomuscular coupling

A number of recent studies have investigated oscillatory activity around 15-30 Hz in the primary motor cortex (M1), both in humans using MEG (Salenius et al., 1997), and in monkeys using local field potential recordings (LFP) (Murthy and Fetz, 1992; Brown et al., 1998; Feige et al., 2000; Kilner et al., 1999, 2000, 2002; Baker et al., 2001, 2003; Jackson et al., 2002). These oscillations appear to be strongest during rest or steady contractions, but may be diminished or modulated during dynamic movements (Kilner et al., 2000; Pfurtscheller and Neuper, 1994; Salmelin and Hari, 1994; Pfurtscheller et al., 1996). Both theoretical and experimental work support that these widespread oscillations are critically related to inhibitory systems and therefore amenable to pharmacological manipulations (Whittington et al., 1995; Wang and Buzsaki, 1996; Pauluis et al., 1999).

That synchronization between two separate neural systems is important for normal functioning is widely accepted in various studies on the sensorimotor system (Bressler et al., 1993; Classen et al., 1998; Grosse et al., 2002, 2003). More recently there has been interest in determining the coupling between ongoing cortical rhythms (using Local Field Potentials, EEG or MEG) and oscillations in the electrical activity of the muscles (measured by surface EMG). Although the function of normal oscillations in the cortex, basal ganglia and cerebellum is far from clear (Farmer, 1998), these different oscillation-frequencies may be important for linking the primary motor cortex and the basal ganglia and cerebellum (Grosse et al., 2002).

The majority of studies investigating neural synchrony or corticomuscular coupling have used ' ' coherence'' as a measure of coupling between ongoing oscillations. Coherence, crudely speaking, is a normalized quantity measuring the degree of time-locked correlation between two signals as a function of frequency. Thus if two noisy waveforms receive a common input of say, 30 Hz, one would expect a peak in their coherence at 30 Hz.

Nevertheless there are a number of technical limitations associated with the current method of measuring the coherence between a single EEG lead and a single rectified EMG lead:

1. It is often assumed that there is temporal stationarity of the EEG and EMG spectra. Previous studies have suggested that EEG/EMG coherence is maximal during sustained contractions and disappears during changing of movements. However, as motor movements are naturally dynamic and non-stationary, it would be desirable to have assays that can track dynamic changes in muscle activity. Newer approaches, such as the wavelet coherence (Lachaux et al., 2002; Saab et al., 2005) may address this issue.

2. The common formulation of coherence relies on pairwise comparisons. Nevertheless the biology suggests that the mapping between the brain and musculature is many-to-many, as opposed to one-to-one (Murthy and Fetz, 1992). Recent analyses on real and simulated data have emphasized that the multivariate approach is much more accurate than pairwise analyses, which can be misleading (Kus et al., 2004).

3. Rectification (taking the absolute value) of the EMG to estimate the envelope. While rectification of the EMG when it clearly consists of individual motor units separated in time may result in the frequency spectrum approaching that of the envelope frequency (Myers et al., 2003), such a situation rarely occurs in practice with surface EMG during anything more than a minimal contraction - the typical scenario. Others have argued that rectification is not warranted on theoretical grounds (for a review, see (Farina et al., 2004)).

Despite these limitations, corticomuscular coherence has proved valuable to investigate different features of the motor system. The effects of sensory input on corticomuscular coupling have been investigated by temporary de-afferentation with ischemia or digital nerve block (Fisher et al., 2002; Pohja and Salenius, 2003). No significant change in the dominant frequencies of the cortico-muscular coherence or EMG-EMG coherence (implying a common cortical signal in both muscles) was detected, suggesting that the sensory feedback loop is not necessary for the generation of corticomuscular coherence.

The 15-30 Hz coherence may be modulated by a number of factors. Kristave-Feige et al. found that corticomuscular coherence was decreased when attention was divided between motor and arithmetic tasks (Kristeva-Feige et al., 2002), but the full extent upon that higher cognitive functions may modulate corticomuscular coherence is still unclear. Pharmacological manipulations that affect the GABA system affect 20 Hz cortical oscillations, but the computed corticomuscular coherence is relatively invariant to pharma-cologic interventions, suggesting that that corticomuscular coherence itself may exhibit homeostasis, and have a functional role in motor control (Baker and Baker, 2003; Salenius et al., 2002).

We prefer the more general term EEG/ EMG ''coupling'' as opposed to ''coherence'', as coherence is a specific mathematical operation which is fundamentally limited to comparing two waveforms. More general techniques that we have employed are able to provide multiple EEG to multiple EMG comparisons (McKeown, 2000; McKeown and Radtke, 2001).

Recently, we have suggested that Empirical Mode Decomposition (EMD) is a way to estimate the envelope of motor unit potentials without having to use rectification (Liao et al., 2005) (Fig. 1). Like ICA, EMD attempts to decompose a time series into

Intrinsic mode functions (IMFs)

Fig. 1. With EMD, the EMG signal is decomposed into components, called intrinsic mode functions (IMFs)

Intrinsic mode functions (IMFs)

Fig. 1. With EMD, the EMG signal is decomposed into components, called intrinsic mode functions (IMFs)

individual components, so that the linear sum of the components approximates the original signal (Huang et al., 1998). However, unlike ICA which examines linear combinations of simultaneous-recorded uni-variate time series, EMD works only on a univariate time series. In EMD, the extracted components are referred to as Intrinsic Mode Functions (IMFs).

IMFs form a complete and 'nearly' orthogonal basis for the original signal. In some

Fig. 2. The motion of the undulating sides of the ''tunnel'' is downward. The ''inflatable'' ring is under the subject's control

"Inflatable" ring

Fig. 2. The motion of the undulating sides of the ''tunnel'' is downward. The ''inflatable'' ring is under the subject's control

Fig. 3. Coherence between different EEG measures and different EMG measures. Note the prominent peak at around 30 Hz for the coherence between the one EEG IC and one IMF. The F3 channel from the EEG had less coherence with the single EMG channel (either with or without rectification). Also note the much reduced coherence between the EEG IC and the surrogate, which had the identical frequency spectrum of the IMF. The p < 0.001 level (determined by

Monte Carlo simulation) is approx 0.02

Fig. 3. Coherence between different EEG measures and different EMG measures. Note the prominent peak at around 30 Hz for the coherence between the one EEG IC and one IMF. The F3 channel from the EEG had less coherence with the single EMG channel (either with or without rectification). Also note the much reduced coherence between the EEG IC and the surrogate, which had the identical frequency spectrum of the IMF. The p < 0.001 level (determined by

Monte Carlo simulation) is approx 0.02

situations, different components may have sections with similar frequencies at different time durations, but locally, any two components will tend to be orthogonal. Because the IMFs are formed by explicitly fitting the envelope of a time series, we posited that IMFs derived from EMG recordings would more closely reflect underlying muscle firing activity, and therefore would better correspond with ongoing brain rhythms (see Fig. 3) (Liao et al., 2005).

Corticomuscular coupling in PD

It is perhaps not surprising that corticomus-cular coherence analysis has been proposed to monitor cortical dysfunction in Parkinson's disease. The normal coherence between MEG signals and simultaneously recorded EMG activity at 15-30 Hz is disrupted in PD patients withdrawn from L-Dopa (Salenius et al., 2002), suggesting a disruption of cortical drive. Coherence between muscles, i.e. EMG-EMG inter-muscular coherence, may indirectly measure cortical influence. Improvements of bradykinesia correspond with increases in EMG-EMG coherence (Brown et al., 2001).

Salenius et al. (2002) found that 3 out of 8 PD patients off L-Dopa medication had abnormally strong MEG-EMG coherence at the much lower frequencies of 5-12 Hz compared with medicated or eight healthy age-matched control subjects. The demonstration that PD patients have a reduction in the corticomuscular coherence at 15-30 Hz, and that L-Dopa improves the coherence in the 15-30Hz bands (Salenius et al., 2002) support the contention that the basal ganglia contribute to the modulation of 15-30 Hz coherence.

One explanation for the presence of these generally non-overlapping frequency bands is that while the higher frequency band (~ 15-30 Hz) represents direct corticomus-cular coupling, the coherent lower frequency band (~10Hz) may represent some temporal aspect of ''online'' motor updating. This has been suggested by Grosse et al., who demonstrated that slow finger movements are actually composed of intermittent (~9Hz) velocity bursts, and on the basis of MEG recordings, suggested that these bursts may be modulated by the cerebello-thalamo-cortical loop (Gross et al., 2002). These intermittent changes in velocity may therefore reflect cerebellar and/or basal ganglia influences on the final motor output (Welsh et al., 1995). We note that the lower frequency band is differentially modulated from the 15-30 Hz band by pharmacological manipulation (Baker and Baker, 2003), suggesting a different mechanism.

Here we suggest a method, based on ICA and Empirical Mode Decomposition (EMD) to infer corticomuscular coupling. We demonstrate that this provides a means to infer ~10Hz oscillations in PD, and that these correlate with UPDRS bradykinesia scores.

Was this article helpful?

0 0
All About Alzheimers

All About Alzheimers

The comprehensive new ebook All About Alzheimers puts everything into perspective. Youll gain insight and awareness into the disease. Learn how to maintain the patients emotional health. Discover tactics you can use to deal with constant life changes. Find out how counselors can help, and when they should intervene. Learn safety precautions that can protect you, your family and your loved one. All About Alzheimers will truly empower you.

Get My Free Ebook

Post a comment