Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
, 143 (2), 582-596

Subthalamic Nucleus Activity Dynamics and Limb Movement Prediction in Parkinson's Disease

Affiliations

Subthalamic Nucleus Activity Dynamics and Limb Movement Prediction in Parkinson's Disease

Saed Khawaldeh et al. Brain.

Abstract

Whilst exaggerated bursts of beta frequency band oscillatory synchronization in the subthalamic nucleus have been associated with motor impairment in Parkinson's disease, a plausible mechanism linking the two phenomena has been lacking. Here we test the hypothesis that increased synchronization denoted by beta bursting might compromise information coding capacity in basal ganglia networks. To this end we recorded local field potential activity in the subthalamic nucleus of 18 patients with Parkinson's disease as they executed cued upper and lower limb movements. We used the accuracy of local field potential-based classification of the limb to be moved on each trial as an index of the information held by the system with respect to intended action. Machine learning using the naïve Bayes conditional probability model was used for classification. Local field potential dynamics allowed accurate prediction of intended movements well ahead of their execution, with an area under the receiver operator characteristic curve of 0.80 ± 0.04 before imperative cues when the demanded action was known ahead of time. The presence of bursts of local field potential activity in the alpha, and even more so, in the beta frequency band significantly compromised the prediction of the limb to be moved. We conclude that low frequency bursts, particularly those in the beta band, restrict the capacity of the basal ganglia system to encode physiologically relevant information about intended actions. The current findings are also important as they suggest that local subthalamic activity may potentially be decoded to enable effector selection, in addition to force control in restorative brain-machine interface applications.

Keywords: Parkinson’s disease; brain computer interface; deep brain recording; machine learning; subthalamic nucleus.

Figures

Figure 1
Figure 1
Experimental setup and analysis pipeline. (A) Deep brain electrode schematic. (B) The directional DBS lead (Boston Scientific). Contacts are distributed along four levels. On levels two and three, there are three segmented contacts (level two: contacts 2/3/4; level three: contacts 5/6/7). (C) The fixed-limb voluntary movement task; upper and lower limb movements performed in separate blocks, with each block preceded with an instruction describing the limb to be moved after hearing the imperative auditory cue. (D) Random-limb voluntary movement task; upper and lower limb movements randomly instructed within the same experimental block. Here the auditory cue prior to each trial also describes the limb to be moved. (E) Flow chart summarizing analysis pipeline. LFP signals are used to predict the limb moved using the naïve Bayes technique. Acc. = acceleration; Ant = anterior; au = arbitrary units; Lat = lateral; Med = medial; Post = posterior.
Figure 2
Figure 2
Localization of all the directional contacts in all 18 subjects included in the study. Distribution of the contacts (blue dots) and the mean coordinate of these contacts (black dots) are shown relative to the STN (grey mesh) in three different planes. In addition, a blue-shaded sphere is shown where the diameter is separately defined for the x, y, and z coordinates and corresponds to a range of 2.5 standard deviations.
Figure 3
Figure 3
Burst detection process. (A) Raw LFP in a 2-s epoch from one of the directional contacts. (B) LFP power in alpha band and detected alpha bursts. Threshold shown by interrupted red line is the 75th percentile power. Note that threshold crossings had to exceed 100 ms in duration to be classified as bursts. (C) LFP power in beta band and detected beta bursts. (D) Detected alpha and beta bursts. (E) Final non-overlapping alpha and beta bursts, and burst-free periods (non-burst). au = arbitrary units.
Figure 4
Figure 4
Classification AUC across various task periods. Task periods included: rest (rst.), pre-cue (-cue.), pre-movement onset (-ons.), and post-movement onset (+ons.). (A) Example of performance of one contralateral hemisphere during fixed-limb blocks. (B) Example of performance of one contralateral hemisphere during random-limb block. (C) Example of performance of one ipsilateral hemisphere during fixed-limb block. (D) Example of performance of one ipsilateral hemisphere during random-limb block. (E) Average across contralateral hemispheres during fixed-limb block (n = 23) and random-limb block (n = 6) across six directional contacts (c2–c7). (F) Average across ipsilateral hemispheres during fixed-limb (n = 13) and random-limb blocks (n = 2) across six directional contacts (c2–c7). A and C are for the left hemisphere of Subject 1 and B and D are for the left hemisphere of Subject 16. Horizontal dashed red line shows the AUC if classification were at chance level.
Figure 5
Figure 5
Burst characteristics during the pre-cue period in the fixed-limb experiment. (A) Example shows power spectral density for the three bursting states (non-burst, alpha, and beta) from the left hemisphere of Subject 1 (channel 4). (B) Raster plot of bursts (top) and mean of beta burst incidence (% of total trials; bottom) at each time point for all trials of fixed-limb blocks (23 hemispheres, 9801 trials in total) aligned to cue (green vertical line). Mean of trial movement onsets is blue vertical line and mean of trial movement offsets is yellow vertical line. (C) Histogram of burst rate for alpha (blue) and beta (red) bursts across all contralateral hemispheres (n = 23). The overlap between the two burst distributions is dark red. (D) Histogram of mean burst duration for alpha and beta bursts across all contralateral hemispheres (n = 23). (E) Histogram of the proportion of the whole pre-cue period taken up by the combined alpha and beta bursts across all contralateral hemispheres (n = 23). Ratio of 0.2 means, for example, that alpha and beta bursts comprise 20% of the pre-cue period. (F) Histogram of proportion of bursts that overlap between alpha and beta across all contralateral hemispheres (n = 23).
Figure 6
Figure 6
Classification AUC during the pre-cue period for different burst conditions [alpha (8–12 Hz), beta (13–30 Hz), and non-burst]. (A) Average across all contralateral hemispheres (n = 23) for same best three contacts (i.e. the three contacts affording the best predictions as measured by the AUC). (B) Average across all contralateral hemispheres (n = 23) for different best three contacts. (C) Average across all ipsilateral hemispheres (n = 13) for same best three contacts. (D) Average across all ipsilateral hemispheres (n = 13) for different best three contacts. ***P < 0.001. Box and whisker plots with median (red horizontal line), 25th and 75th percentiles (bottom and top edges of the box, respectively), and outliers (red+) noted. Horizontal dashed red line shows the AUC if classification were at chance level.
Figure 7
Figure 7
Feature importance for prediction during different task periods in fixed-limb blocks for raw LFP before classification into burst conditions. (A) Average across all contralateral hemispheres (n = 23) for the same best three contacts in different task periods. Feature importance was correlated across task periods e.g. between pre-cue and pre-movement onset, Spearman’s rho = 0.833, P = 0.008, between pre-cue and post-movement onset, Spearman’s rho = 0.933, P < 0.001, and between pre- and post-movement onset, Spearman’s rho = 0.85, P = 0.006. (B) Average across all contralateral hemispheres (n = 23) for the different best three contacts. Feature importance was correlated across task periods e.g. between pre-cue and pre-movement onset, Spearman’s rho = 0.87, P = 0.003, between pre-cue and post-movement onset, Spearman’s rho = 0.862, P = 0.004, and between pre- and post-movement onset, Spearman’s rho = 0.917, P = 0.0013. (C) Average across all ipsilateral hemispheres (n = 13) for the same best three contacts. Feature importance was correlated across task periods e.g. between pre-cue and pre-movement onset, Spearman’s rho = 0.783, P = 0.0172, between pre-cue and post-movement onset, Spearman’s rho = 0.767, P = 0.021, and between pre- and post-movement onset, Spearman’s rho = 0.467, P = 0.2125. (D) Average across all ipsilateral hemispheres (n = 13) for the different best three contacts. Feature importance was correlated across task periods e.g. between pre-cue and pre-movement onset, Spearman’s rho = 0.967, P < 0.001, between pre-cue and post-movement onset, Spearman’s rho = 0.967, P < 0.001, and between pre- and post-movement onset, Spearman’s rho= 0.95, P < 0.001. Task periods were pre-cue (-cue.), pre-movement onset (-ons.), and post-movement onset (+ons.).
Figure 8
Figure 8
Feature importance during the pre-cue period for the burst conditions: non-bursts, alpha bursts, and beta bursts. (A) Average across all contralateral hemispheres (n = 23) for the same best three contacts. (B) Average across all contralateral hemispheres (n = 23) for the different best three contacts. (C) Average across all ipsilateral hemispheres (n = 13) for the same best three contacts. (D) Average across all ipsilateral hemispheres (n = 13) for the different best three contacts.

Similar articles

See all similar articles

References

    1. Androulidakis AG, Kühn AA, Chen CC, Blomstedt P, Kempf F, Kupsch A., et al. Dopaminergic therapy promotes lateralized motor activity in the subthalamic area in Parkinson's disease. Brain 2007; 130: 457–68. - PubMed
    1. Anidi C, O'Day JJ, Anderson RW, Afzal MF, Syrkin-Nikolau J, Velisar A., et al. Neuromodulation targets pathological not physiological beta bursts during gait in Parkinson's disease. Neurobiol Dis 2018; 120: 107–17. - PMC - PubMed
    1. Avants BB, Epstein CL, Grossman M, Gee JC. Symmetric diffeomorphic image registration with cross-correlation: evaluating automated labeling of elderly and neurodegenerative brain. Med Image Anal 2008; 12: 26–41. - PMC - PubMed
    1. Brittain JS, Brown P. Oscillations and the basal ganglia: motor control and beyond. Neuroimage 2014; 85: 637–47. - PMC - PubMed
    1. Brown P, Oliviero A, Mazzone P, Insola A, Tonali P, Di Lazzaro V. Dopamine dependency of oscillations between subthalamic nucleus and pallidum in Parkinson's disease. J Neurosci 2001; 21: 1033–8. - PMC - PubMed
Feedback