Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Sep 1.
Published in final edited form as: J Neurosci Methods. 2020 Jun 18;343:108811. doi: 10.1016/j.jneumeth.2020.108811

A Novel Method for Calculating Beta Band Burst Durations in Parkinson’s Disease Using a Physiological Baseline

RW Anderson 1, YM Kehnemouyi 1, RS Neuville 1,a, KB Wilkins 1, CM Anidi 1,b, MN Petrucci 1, JE Parker 1, A Velisar 1,c, H Bronte-Stewart 1,2
PMCID: PMC7437353  NIHMSID: NIHMS1609153  PMID: 32565222

Abstract

Background

Pathologically prolonged bursts of neural activity in the 8–30 Hz frequency range in Parkinson’s disease have been measured using high power event detector thresholds.

New Method

This study introduces a novel method for determining beta bursts using a power baseline based on spectral activity that overlapped a simulated 1/f spectrum. We used resting state local field potentials from people with Parkinson’s disease and a simulated 1/f signal to measure beta burst durations, to demonstrate how tuning parameters (i.e., bandwidth and center frequency) affect burst durations, to compare burst duration distributions with high power threshold methods, and to study the effect of increasing neurostimulation intensities on burst duration.

Results

The baseline method captured a broad distribution of resting state beta band burst durations. Mean beta band burst durations were significantly shorter on compared to off neurostimulation (p = 0.0046), and their distribution shifted towards that of the 1/f spectrum during increasing intensities of stimulation.

Comparison with Existing Methods

High power event detection methods, measure duration of higher power bursts and omit portions of the neural signal. The baseline method captured the broadest distribution of burst durations and was more sensitive than high power detection methods in demonstrating the effect of neurostimulation on beta burst duration.

Conclusions

The baseline method captured a broad range of fluctuations in beta band neural activity and demonstrated that subthalamic neurostimulation shortened burst durations in a dose (intensity) dependent manner, suggesting that beta burst duration is a useful control variable for closed loop algorithms.

Keywords: Beta fluctuations, burst durations, local field potentials, thresholding, Parkinson’s disease, Deep Brain Stimulation

1. Introduction

Local field potentials (LFPs) represent the summed electrical activity of local neuronal networks and have been represented classically by power spectral density (PSD) analyses and spectrograms. The PSD represents the average power of each frequency component but not the exact timing of the events that generate it. Physiological broadband neural activity is represented by a 1/f spectrum on a PSD (Gao et al., 2017; He, 2014; Manning et al., 2009; Shreve et al., 2017). This 1/f spectrum strongly correlates with single-neuron spiking in humans and has been shown to be related to excitation and inhibition (Manning et al., 2009). Physiological neural activity consists of short duration (40 – 120 ms) neuronal oscillations and synchrony that represent normal signal processing in the sensorimotor network (Courtemanche et al., 2003; Feingold et al., 2015; Murthy and Fetz, 1996, 1992). Neuronal oscillations appearing as peaks above the broadband 1/f spectrum within the alpha/beta band (8 – 30 Hz), represents an averaged measure of a state of higher probability of exaggerated neuronal oscillations and neural synchrony, which are pathophysiological markers of the hypokinetic state in Parkinson’s disease (PD) (Anidi et al., 2018; Bergman et al., 1994; Brown, 2003; Deffains et al., 2018; Hammond et al., 2007; Kühn et al., 2008, 2006; Velisar et al., 2019; Whitmer et al., 2012).

The classical tool used to measure how neural signals change with time is the spectrogram, which is made up of consecutive, stepped and overlapping short PSDs. Spectrograms, however, have a fixed time-frequency resolution: longer time windows result in better frequency resolution but poor temporal resolution, while short time windows allow temporal events to be easily distinguished but have poor frequency distinction. There is an emerging need to develop methods for analyzing neural signals over time that do not compromise frequency or temporal resolution.

This became pertinent for PD investigations since Feingold et al. suggested that the durations of beta band power fluctuations may be the key difference between physiological and PD related (pathological) neurophysiology (Feingold et al., 2015). Current published methods to measure the dynamics of beta band oscillatory activity have focused on event detection of a high power burst: aspects of the signal were only included in the burst analysis if and when instantaneous power rose above a threshold of the power of a bandpass filtered signal (Cagnan et al., 2019; Deffains et al., 2018; Feingold et al., 2015; Lofredi et al., 2019; Meidahl et al., 2019; Tinkhauser et al., 2018, 2017a, 2017b; Torrecillos et al., 2018). More recently, it has been shown that burst duration, rather than burst power, was a more reliable metric to discriminate healthy from parkinsonian neural activity in non-human primates (Deffains et al., 2018).

In this study we introduce a new method to measure fluctuations in LFP power using a baseline of power. The baseline was calculated from a frequency band within the same PD LFP spectrum, which overlapped and had similar burst dynamics to the simulated physiological 1/f spectrum. This baseline, from an ‘inert’ band in the same spectrum, can be used to determine burst durations and power in any other band in the spectrum and this allows comparison of burst dynamics across participants, as commonly used for LFP power (Anidi et al., 2018; Syrkin-Nikolau et al., 2017). We demonstrate that the baseline method captured a broader range of burst durations (with no power discrimination) in the pathological beta band compared to other power threshold methods. The baseline method was used to investigate the effect of increasing intensities of STN DBS on resting state beta band burst duration distributions.

2. Methods

2.1. Human subject data

Nine PD participants (six male, fifteen STNs) had bilateral implantation of DBS leads (model 3389, Medtronic, Inc.) in the sensorimotor region of the STN using a standard functional frameless stereotactic technique and multipass microelectrode recording (MER) (Brontë-Stewart et al., 2010; Quinn et al., 2015). Dorsal and ventral borders of each STN were determined during MER, and the base of electrode zero was placed at the ventral border of the STN. The leads were connected through Medtronic research extensions (model 3708760) to the implanted investigative neurostimulator (Activa® PC+S, Medtronic, Inc. FDA Investigator Device Exemption approved). The preoperative selection criteria, surgical technique, and assessment of the participants have been previously described (Brontë-Stewart et al., 2010; Quinn et al., 2015). The participants gave written informed consent to participate in the study, which was approved by the Food and Drug Administration (FDA) and the Stanford School of Medicine Institutional Review Board (IRB). Long-acting dopaminergic medication was withdrawn over 24 h (72 h for extended-release dopamine agonists), and short-acting medication was withdrawn over 12 h before the study visit. Participants had a mean age of 55.3 ± 9.5 years at the time of DBS titration experiments with an average disease duration of 8.9 ± 3.3 years. Participants had DBS surgery an average 2.4 years before experiments were conducted. Resting state data from one participant in this cohort was collected 1-year post DBS surgery.

2.2. Experimental protocol

Recordings were collected in the Stanford Human Motor Control and Neuromodulation laboratory. Experiments were performed off medication and the participants were instructed to remain seated and as still as possible with their eyes open during all recordings. All resting state off stimulation data was collected at least 60 minutes after DBS was turned off (Trager et al., 2016). Stimulation titration data were collected using randomized presentations of 0%, 25%, 50%, 75% and 100% of Vmax, which was the clinically equivalent DBS intensity: the intensity using a single active contact that represented the intensity being used using one or multiple contacts for clinical DBS. The last 30 seconds of a 60 second resting period at the respective stimulation intensity was used for the analysis.

2.3. Data acquisition and analysis

STN LFPs were recorded from electrode contact pair 0–2 (13 STNs) or 1–3 (3 STNs) on the DBS lead with LFPs high pass filtered at 0.5 Hz, low pass filtered at 100 Hz, amplified at gains set to minimize stimulation artifact and sampled at 422 Hz (10-bit resolution). Stimulation for titration experiments (60 μS pulse width, 140 Hz) were delivered through contact 1 or 2 with the maximum stimulation intensity set to the clinically equivalent single contact voltage. The power spectral density (PSD) estimate was calculated using Welch’s method with a 1-second sliding Hanning window with 50% overlap (Welch, 1967). All analysis was conducted in MATLAB (version 9.6, The MathWorks Inc. Natick, MA, USA).

2.4. Simulation of “physiological” neural activity

Simulation of local field potential activity in a normal brain was performed using pink (1/f) noise, generated using the MATLAB routine dsp.ColoredNoise in a period of 36 seconds, with a sample frequency of 422 samples/second and the amplitude matched to the roll off of the in vivo human data. This theoretically simulates a broadband spectrum, which is that seen in non-oscillating, physiological neural activity in the normal, non-Parkinsonian state. Each burst analysis method was performed on the middle 30 seconds of simulated data to eliminate filter edge effects and the simulation was repeated if any burst methods failed to find bursts in the simulated data. This sequence was repeated until 1000 episodes of simulated data with identified bursts were acquired. Filter bandwidths were tested from 1Hz-17Hz and center frequencies from 10 Hz to 24 Hz to encompass the bandwidths and center frequencies previously published in the literature.

2.5. Statistics

Burst duration and patient demographics are presented as mean +/− standard deviation. Comparison of burst duration between gamma band PD and pink noise in Figure 2 are calculated using a two-sample t-test. Burst durations calculated between off stimulation and 100% stimulation in Figure 5 were compared using a paired sample t-test for each method.

Figure 2:

Figure 2:

Open in a new tab

Bandpass filtered PD LFP and pink noise filtered for the gamma frequency band (CF = 54 Hz, BW = 6Hz). Parkinson’s gamma band data (A) and simulated pink noise data (B) are shown using the same axes. LFP power (black), identified troughs (red dots), thresholds (red line) and identified bursts using the Anderson method in PD (C) and pink noise (D). Histograms of the associated burst durations for PD (E) and pink noise (F) are similar.

Figure 5:

Figure 5:

Open in a new tab

Beta power and burst durations are suppressed by deep brain stimulation. (A) Five different stimulation conditions (0%, 25%, 50%, 75% and 100% of clinical equivalent stimulation) are compared with simulated 1/f pink noise in a representative individual. Vertical red bars, a subset of the beta frequency band used for analysis in B that is highly dependent on stimulation intensity. (B) Burst durations calculated for the same 30 second windows are decreased with increasing levels of stimulation and are driven towards the normal burst duration expected with simulated pink noise as calculated by the baseline method. Mean burst durations are shown by labeled red bars. Boxplots of mean burst durations demonstrating the effect of deep brain stimulation on mean beta burst durations as calculated by the: C, baseline; D, high power detection/lower threshold; E, high power detection/threshold at 0%, 25%, 50%, 75%, and 100% Vmax. The thick black line represents the median and the open circle represents the group mean.

3. Results

The Baseline Method – Determining Burst Durations Using a Physiological Baseline

3.1. Establishing the physiological baseline

The goal of the baseline method was to capture all possible durations of fluctuations (bursts) in power by determining a baseline of power, using an inert, low power band in the PD spectrum, which overlapped the physiological 1/f signal. Figure 1 demonstrates the similarities and differences between a PD spectrum and a simulated 1/f spectrum modeled as pink noise. The PD participant’s raw unfiltered LFP had elevated amplitude when compared to the pink noise data (Fig. 1A and B).

Figure 1:

Figure 1:

Open in a new tab

Local field potentials from a Parkinsonian patient are compared to simulated 1/f pink noise matched to the same roll off. Broad spectrum Parkinsonian local field potentials (A) are larger in amplitude than the simulated pink noise (B). Both signals are plotted on the same PSD (C), the power in the beta (13–30Hz) frequency range is elevated when compared with the gamma frequency range (40–70Hz). (D) Sub bands of the Parkinson’s and pink noise gamma (blue lines) and beta (green lines) from C are band pass filtered showing the elevation of power in the beta frequency range (Bottom Right).

When both signals were plotted on the same PSD, Fig. 1C, the power in a low gamma frequency band (40 – 70 Hz) of the PD spectrum overlapped that of the simulated physiological 1/f spectrum, whereas the power in the alpha and beta (8 – 30 Hz) frequency range was elevated above the 1/f spectrum. The LFPs from the 1/f and PD signals were each filtered in two 6 Hz bands: one in the band where the two PSDs overlapped (51 – 57 Hz, blue lines, Fig. 1C) and the other around the beta peak of the PSD spectrum (15.13 – 21.13 Hz, green vertical lines, Fig. 1C). In the overlapping band the PD and 1/f filtered signals were similar in amplitude, Fig. 1D top panels, however, the PD signal had greater amplitude than that in the 1/f signal in the elevated beta sub-band, Fig. 1D lower panels.

3.2. Burst durations were similar in the PD and 1/f overlapping bands, establishing a physiological baseline from the PD spectrum

Figure 2A and B demonstrate the bandpass filtered signal from an overlapping band of the PD and 1/f spectrum, respectively. The bandpass filtered signals were squared and the envelopes of those signals are shown in Figure 2C and D respectively.

A trough detection algorithm identified the local minima of the envelopes of the filtered squared signals (red dots, Figs. 2C and D). Troughs represented the local smallest peak amplitudes of the filtered squared signal, and troughs of the low gamma band were used to reject the effect of power elevations or oscillatory activity captured as peaks above the PD gamma band. To derive benefits from averaging and to conserve the energy in the band, five overlapping 6-Hz bands centers at 48, 51, 54, 57 and 60 Hz were constructed and the median trough values from each band calculated and averaged. Multiples of the average median trough power were considered in determining the threshold in order to reliably detect gamma power above the estimated device noise floor of the Activa™ PC+S neurostimulator (Anidi et al., 2018). It was determined that thresholds ranging from 3 to 5 times the median trough power were conservatively above the device noise power, while still being low enough to reflect the desire for a baseline of power. These three thresholds were evaluated in their effect on capturing bursts clearly and calculating mean burst durations across different stimulation intensities. All three thresholds for the baseline method demonstrated a significant difference between mean burst duration during no (0%) versus 100% Vmax DBS with minimal differences seen between thresholds (see Supplementary Information, Figure S1). The threshold of 4 times the median trough power was selected as a reliable baseline going forward. The red line in Figs. 2C and D represents 4 times the median power of the troughs, which formed the threshold from which burst durations and power were calculated for each signal. The duration of a fluctuation or burst of power was defined as the period between ‘zero’ crossings of the envelope across the baseline; the mean power of the burst was the average power of the filtered and squared LFP signal bounded by the start and end time of the burst. The histograms in Fig. 2E and F demonstrate that the distributions and respective means of burst durations were similar between the overlapping PD and 1/f bands. The mean burst durations were 181 ± 102 ms and 153 ± 76 ms, respectively, t(109)=1.68; p=0.096, which were similar to those reported for physiological neural activity in non-human primates (40 – 120 ms, Feingold 2015). This supported the hypothesis that this non-oscillatory portion of the PD spectrum comprised largely short duration (physiological) fluctuations of neural activity; we therefore designated the baseline of the PD overlapping band as the ‘physiological’ baseline that would be used to calculate burst dynamics or fluctuations in pathological or elevated bands of the PD spectrum.

3.3. Calculation of burst durations depend on the choice of center frequency and bandwidth

Figure 3 demonstrates how varying filter bandwidth and center frequency affect burst duration. Mean burst durations of the 1000 episodes of pink noise were calculated with the signal filtered in successive bandwidth ranges from 1 – 10 Hz at each of the center frequencies of 10 Hz, 15 Hz, and 20 Hz.

Figure 3:

Figure 3:

Open in a new tab

Burst durations are calculated for the baseline method using 1000 runs of simulated pink noise. The effects of bandwidth and center frequency (10, 15, and 20 Hz) are shown. Error bars represent standard deviations of the mean.

The mean burst duration of the 1/f spectrum depended on the bandwidth of the filter used: it increased exponentially for narrow bandwidths and followed an asymptote towards little change at bandwidths longer than 6 Hz. At bandwidths greater than 2 Hz, mean burst durations of pink noise were higher at a center frequency of 10 Hz compared to 20 Hz. At bandwidths greater than 6 Hz, the mean burst durations were greater at the 10 Hz center frequency than at both 15 Hz and 20 Hz. Burst durations asymptote after 6 Hz, suggesting that larger bandwidths would not be expected to significantly alter burst durations.

3.4. The baseline method captured a broader distribution of burst durations compared to methods using a threshold of power

Figure 4 demonstrates the differences in beta burst duration distribution when calculated using the baseline method (Fig. 4B, E, H), compared to two different high power event detection methods Fig. 4C, F, I, and Fig. 4D, G, J, respectively (Feingold et al., 2015; Tinkhauser et al., 2017a). The same thirty seconds of a PD LFP signal was used for all the methods and was filtered using a zero-phase 8th order Butterworth bandpass filter with a constant 6Hz bandwidth around the same center frequency (18.1 Hz). The filtered signal is shown in Fig. 4A.

Figure 4:

Figure 4:

Open in a new tab

Comparison of the three burst duration methods (Baseline, High Power Detection Lower Threshold, High Power Detection/Threshold) processed with the same center frequency (18.1Hz) and bandwidth (6 Hz). (A) Bandpass filtered LFP used for each analysis. Squared LFP and power envelope for the Baseline (B, E), High Power Detection Lower Threshold (C, F), and High Power Detection/Threshold (D, G) methods. Histograms of the resulting burst durations for the Baseline (H), High Power Detection Lower Threshold (I) and High Power Detection/Threshold (J) methods. Red line indicates the threshold for burst quantification (E, F, G) and the blue line represents the threshold for burst selection (F).

For the baseline method, the bandpass filtered signal was squared (Fig. 4B), and an amplitude envelope was created by linearly connecting consecutive peaks of the filtered and squared LFP signal to form an envelope of the maximum power, (Fig. 4E). The baseline was established by averaging the median trough amplitudes from 5 consecutive overlapping 6 Hz bands in the 45–63 Hz PD gamma spectrum. The median of the troughs (minima) of each episode of the envelope power were calculated and the average median trough power was determined; this was the baseline from which fluctuations of power in the band in the PD spectrum would be measured. In the high power detection, lower threshold method the same bandpass filtered signal was squared, Fig. 4C, and then smoothed using a Hanning window moving average filter, Fig. 4F. The high power detection, lower threshold method identified a burst when the signal power rose above a power threshold that was 3 times the median power of the signal, blue line Fig. 4F, and used a lower power threshold of 1.5 times the median power to calculate the burst durations, red line Fig. 4F (Feingold et al., 2015). The high power detection/threshold method rectified the signal but did not square it, Fig. 4D; it then smoothed the rectified signal using a 400 ms rectangular moving average filter, Fig. 4G, and identified and classified a burst using a power threshold, which could range from the 55th to the 95th percentile of the power in the smoothed signal (75th percentile shown, (Tinkhauser et al., 2017a)).

In all methods, the duration of a burst was calculated as the interval between successive crossings of the signal over the chosen baseline or threshold (red line) and portions of the neural signal below the red line were omitted from analysis. Fig. 4H, I, J demonstrate how the distributions of burst durations from the same filtered signal differed, using the three methods. The baseline method resulted in the broadest distribution of beta burst durations (mean 1545 ± 1001 ms, range 405–3547 ms), the high power detection lower threshold method had the narrowest distribution (mean 264 ± 65 ms, range 192–386 ms), and the high power detection/threshold method had a distribution in between the other methods (mean 439 ± 292 ms, range 147–1431 ms). The baseline method differed in its definition of a burst from the other two methods, specifically in the intent to measure all fluctuations of power over a baseline rather than only the durations of higher power fluctuations using an event detection threshold, that was a percentage of the overall power of the spectrum.

3.5. Resting state beta burst durations decrease during STN DBS at increasing intensity

The baseline, high power detection, lower threshold method and high power detection/threshold methods was used to investigate the effect of increasing the intensity of STN DBS on resting state beta band burst durations, Figure 5.

Figure 5A demonstrates the effect of randomized epochs of STN DBS at different intensities on a resting state PSD from a representative individual with PD; the pink noise estimate of the 1/f PSD is demonstrated in black. The PD PSDs demonstrate attenuation of the resting state beta band power during STN DBS at increasing intensities from 0% (no DBS, red PSD, Fig. 5A) to 100% of Vmax (3.7 V, light blue PSD, Fig. 5A). Burst durations of the six Hz band most modulated by increasing intensities of DBS (red vertical lines, Fig. 5B) were calculated using the baseline method. In this individual there was a broad distribution of resting state burst durations at no (0%) DBS, with a mean burst duration of 1914 ± 1833 ms, which increased along with minimal beta band power attenuation at a small increase in intensity (25% of Vmax). As the DBS intensity increased above 50% of Vmax, the distribution of beta band burst durations shifted towards shorter burst durations and mean beta burst duration decreased. Fig. 5C, D and E and Table 1 demonstrate the effect of increasing intensity of DBS on burst durations for the full cohort of 15 STNs, using the baseline, high power detection, lower threshold method and high power detection/threshold methods respectively. The baseline method showed the strongest change in burst duration between 0% Vmax (no DBS) and 100% Vmax: (t(14)=3.36, p=0.0046), compared to the high power detection, lower threshold method (t(14)=2.74, p=0.016), and the high power detection/threshold method (t(14)=2.70, p=0.017), both of which only measured durations of high power bursts.

4. Discussion

In this study, we introduce a novel method for measuring LFP power fluctuations (bursts) using a baseline of power rather than a high power event detection threshold. The baseline was calculated from the troughs of power of a frequency band within the same PD STN spectrum, which overlapped the simulated physiological 1/f spectrum, and which had similar burst dynamics to physiological neural activity. The baseline was then used to calculate the burst durations in the elevated beta band of interest. Factors that influence burst duration and power calculation include the bandwidth and center frequency of the bandpass filter, the method of rectification (absolute value or squaring of the signal) and the type of smoothing applied to the signal. The baseline method was designed with the intent to detect both low and high power fluctuations in the PD pathological band of interest in order to investigate the effect of increasing intensities of STN DBS on resting state beta band burst durations. There was a significant reduction in mean beta band burst duration at the clinically equivalent DBS intensity compared to no DBS. The distribution of and mean burst durations shifted towards that of shorter durations as the DBS intensity increased. As short beta burst durations have been associated with physiological neural activity (Deffains et al., 2018; Feingold et al., 2015; Murthy and Fetz, 1996, 1992), these results suggest that the effect of DBS is to restore beta band oscillatory activity in PD towards that of the physiological state (He et al., 2014).

4.1. Measurements of the dynamics of pathological beta band neural activity in Parkinson’s disease

Local field potentials (LFPs) represent the summed electrical activity of local neuronal networks and are classically represented by power spectral density (PSD) analysis and spectrograms, which do not have adequate temporal and frequency resolution to accurately measure the temporal dynamics of neural activity. Several new methods for performing spectral deconstruction have emerged that utilize wavelets or bandpass filtering of local field potentials. These methods help to quantify changes in spectral power over time in narrower frequency bands, which has improved the ability to quantify the temporal nature of neuronal oscillatory activity. Currently, the published methods to measure the dynamics of beta band oscillatory activity have focused on event detection to define high power bursts. Aspects of the signal were only included in the burst analysis if and when instantaneous power rose above a threshold of the power of the LFP signal within that band or within the overall beta band, with or without a temporal threshold (e.g. power elevated for greater than a defined number of oscillatory cycles), (Cagnan et al., 2019; Deffains et al., 2018; Feingold et al., 2015; Lofredi et al., 2019; Meidahl et al., 2019; Tinkhauser et al., 2018, 2017a, 2017b; Torrecillos et al., 2018). Most publications have followed the high power detection/threshold method, which detected a burst when power exceeded a threshold of 75% of the power of the rectified smoothed signal. Activity below this power was discarded. In contrast, the baseline method introduced in this study used a baseline power determined from a band in the PD spectrum that overlapped the 1/f signal, and whose burst dynamics resembled that of physiological neural activity. In this way, we sought to include as much of the signal from the band of interest as possible, including lower power fluctuations or bursts. This resulted in a broader range of burst durations compared to either of the high power detection methods. The comparison of the baseline and high power detection lower threshold methods was more straightforward, as the signal was not as highly smoothed as that of high power detection/threshold method. The high power detection lower threshold method resulted in detection of high power bursts, mainly of a shorter duration (Fig. 4I), whereas the lower power baseline of the baseline method resulted in the inclusion of high and lower power fluctuations, with a broader distribution of durations (Fig. 4H). Considering that the different methods highlight different aspects of beta dynamics, it is likely that the optimal method will rely on the question at hand. Deffains et al. demonstrated that STN beta burst duration, as calculated by a threshold method, was a more reliable metric than power to discriminate healthy and parkinsonian neural activity in non-human primates (Deffains et al., 2018).

4.2. Burst duration varied with the choice of bandwidth, center frequency, and smoothing filters

Evaluation of 1000 samples of simulated 1/f neural activity filtered using different bandwidths at three different beta band center frequencies revealed that the mean burst duration had an inverse exponential relationship with bandwidth: at bandwidths shorter than 5 Hz, the mean burst duration progressively increased. Mean burst durations were shorter if the center frequency was 20 Hz compared to 10 Hz. This suggests that burst analysis methods should attempt to use the same bandwidth across each frequency band and take into account different center frequencies if applicable. It is evident that the method used to rectify and/or smooth the raw LFP signal will affect the calculated burst duration and power (Schmidt et al., 2020).

4.3. Beta burst duration as a biomarker of the efficacy of therapy and of Parkinsonian motor and gait impairment

In this study, we demonstrated that the mean resting state beta band burst duration and burst duration distributions progressively decreased with increasing intensities of DBS. Prolonged resting state beta burst durations have been correlated with with gait impairment and freezing of gait (FOG) (Anidi et al., 2018). In that study both 60 Hz and 140 Hz STN DBS decreased mean beta burst duration in freezers, and improved FOG.

The results of this study and a growing body of evidence support the hypothesis that longer duration beta bursts in subcortical structures, equating to longer periods of beta oscillations and synchrony, represent pathological or impaired sensorimotor processing in Parkinson’s disease (Anidi et al., 2018; Cagnan et al., 2019; Deffains et al., 2018; Lofredi et al., 2019; Tinkhauser et al., 2018, 2017a, 2017b) and that one mechanism of STN DBS is to remove pathological sensorimotor network processing, while leaving intact normal physiological processing (Anidi et al., 2018; Holt et al., 2019; Tinkhauser et al., 2017a; Torrecillos et al., 2018). The discovery that averaged beta band power was attenuated during increasing intensities of DBS led to its successful use as the control variable in closed loop DBS studies (Afzal et al., 2019; Eusebio et al., 2011; Little et al., 2016a, 2016b, 2013; Piña-Fuentes et al., 2019, 2017; Rosa et al., 2017, 2015; Syrkin-Nikolau et al., 2017; Velisar et al., 2019; Whitmer et al., 2012). Similarly, the accumulating evidence suggests that beta burst duration will be a functionally relevant, patient specific control variable for closed loop DBS.

5. Conclusions

The baseline method is a novel method for characterizing low and high power fluctuations of activity in local field potentials. This novel method tolerates a wide range of brain activities, as it can be used to calculate bursts in a pathological band of interest from any power spectrum which exhibits behavior similar to a 1/f distribution. We validated this method using pathological activity from people with Parkinson’s disease and simulated physiological neural activity. The baseline was determined from a band within the PD spectrum, which overlapped with the simulated 1/f physiological spectrum and had similar burst duration distributions and mean burst durations. Burst durations calculated using the baseline method captured a wide range of Parkinsonian spectral activity: it revealed mainly short burst durations in the inert band and captured the longest and broadest distribution of burst durations in a pathological beta band compared to two high power event or burst detection methods. The baseline method demonstrated that increased intensities of subthalamic neurostimulation shortened mean beta band burst durations and shifted their distribution towards that of the physiological spectrum. We suggest that this novel method is well suited to quantify the full range of fluctuations in beta band neural activity in the PD brain and is applicable to other bands of interest for other neuropsychological diseases. Beta band burst duration is a functionally relevant, patient specific control variable that can be used for closed loop neuromodulation algorithms in Parkinson’s disease.

Supplementary Material

1

Highlights.

  • A novel method for measuring fluctuations in subthalamic local field potential power in Parkinson’s disease using a baseline of power.

  • Modeling physiological brain activity using a simulated 1/f spectrum.

  • Burst durations depend on choice of bandwidth and center frequency.

  • Burst durations progressively shortened during increasing intensities of neurostimulation.

  • Comparison of the baseline to high power event detection methods in the distribution of beta burst durations and the effect of increasing intensities of neurostimulation

Acknowledgements

We would like to thank the members of the Human Motor Control and Neuromodulation Lab and the participants in the study who without their help, none of this would be possible.

Funding

This work is supported in part by grant number PF-FBS-1899 from the Parkinson’s Foundation. Additional funding was provided by the NIH Brain Initiative 1UH3NS107709, NINDS Grant 5 R21 NS096398-02, Michael J. Fox Foundation (9605), NIH Grant AA023165-01A1, Robert and Ruth Halperin Foundation, The Sanches Family Foundation, John A. Blume Foundation and the Helen M. Cahill Award for Research in Parkinson’s Disease and Medtronic Inc. provided devices but no financial support.

Footnotes

Declaration of Competing Interest

H.B.S. is a member of the Medtronic Inc. Clinical Advisory Board

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Afzal MF, Velisar A, Anidi C, Neuville R, Prabhakar V, Bronte-Stewart H, 2019. Proceedings #61: Subthalamic Neural Closed-loop Deep Brain Stimulation for Bradykinesia in Parkinson’s Disease. Brain Stimulation 12, e152–e154. 10.1016/j.brs.2019.03.019 [DOI] [Google Scholar]
  2. Anidi C, O’Day JJ, Anderson RW, Afzal MF, Syrkin-Nikolau J, Velisar A, Bronte-Stewart HM, 2018. Neuromodulation targets pathological not physiological beta bursts during gait in Parkinson’s disease. Neurobiol. Dis. 120, 107–117. 10.1016/j.nbd.2018.09.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Bergman H, Wichmann T, Karmon B, DeLong MR, 1994. The primate subthalamic nucleus. II. Neuronal activity in the MPTP model of parkinsonism. J. Neurophysiol. 72, 507–520. 10.1152/jn.1994.72.2.507 [DOI] [PubMed] [Google Scholar]
  4. Brontë-Stewart H, Louie S, Batya S, Henderson JM, 2010. Clinical motor outcome of bilateral subthalamic nucleus deep-brain stimulation for Parkinson’s disease using image-guided frameless stereotaxy. Neurosurgery 67, 1088–1093; discussion 1093. 10.1227/NEU.0b013e3181ecc887 [DOI] [PubMed] [Google Scholar]
  5. Brown P, 2003. Oscillatory nature of human basal ganglia activity: relationship to the pathophysiology of Parkinson’s disease. Mov. Disord. 18, 357–363. 10.1002/mds.10358 [DOI] [PubMed] [Google Scholar]
  6. Cagnan H, Mallet N, Moll CKE, Gulberti A, Holt AB, Westphal M, Gerloff C, Engel AK, Hamel W, Magill PJ, Brown P, Sharott A, 2019. Temporal evolution of beta bursts in the parkinsonian cortical and basal ganglia network. Proc. Natl. Acad. Sci. U.S.A. 116, 16095–16104. 10.1073/pnas.1819975116 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Courtemanche R, Fujii N, Graybiel AM, 2003. Synchronous, focally modulated beta-band oscillations characterize local field potential activity in the striatum of awake behaving monkeys. J. Neurosci. 23, 11741–11752. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Deffains M, Iskhakova L, Katabi S, Israel Z, Bergman H, 2018. Longer β oscillatory episodes reliably identify pathological subthalamic activity in Parkinsonism. Mov. Disord. 33, 1609–1618. 10.1002/mds.27418 [DOI] [PubMed] [Google Scholar]
  9. Eusebio A, Thevathasan W, Doyle Gaynor L, Pogosyan A, Bye E, Foltynie T, Zrinzo L, Ashkan K, Aziz T, Brown P, 2011. Deep brain stimulation can suppress pathological synchronisation in parkinsonian patients. J. Neurol. Neurosurg. Psychiatry 82, 569–573. 10.1136/jnnp.2010.217489 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Feingold J, Gibson DJ, DePasquale B, Graybiel AM, 2015. Bursts of beta oscillation differentiate postperformance activity in the striatum and motor cortex of monkeys performing movement tasks. PNAS 112, 13687–13692. 10.1073/pnas.1517629112 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Gao R, Peterson EJ, Voytek B, 2017. Inferring synaptic excitation/inhibition balance from field potentials. Neuroimage 158, 70–78. 10.1016/j.neuroimage.2017.06.078 [DOI] [PubMed] [Google Scholar]
  12. Hammond C, Bergman H, Brown P, 2007. Pathological synchronization in Parkinson’s disease: networks, models and treatments. Trends Neurosci. 30, 357–364. 10.1016/j.tins.2007.05.004 [DOI] [PubMed] [Google Scholar]
  13. He BJ, 2014. Scale-free brain activity: past, present, and future. Trends Cogn. Sci. (Regul. Ed.) 18, 480–487. 10.1016/j.tics.2014.04.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Holt AB, Kormann E, Gulberti A, Pötter-Nerger M, McNamara CG, Cagnan H, Baaske MK, Little S, Köppen JA, Buhmann C, Westphal M, Gerloff C, Engel AK, Brown P, Hamel W, Moll CKE, Sharott A, 2019. Phase-Dependent Suppression of Beta Oscillations in Parkinson’s Disease Patients. J. Neurosci. 39, 1119–1134. 10.1523/JNEUROSCI.1913-18.2018 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Kühn AA, Kempf F, Brücke C, Gaynor Doyle L, Martinez-Torres I, Pogosyan A, Trottenberg T, Kupsch A, Schneider G-H, Hariz MI, Vandenberghe W, Nuttin B, Brown P, 2008. High-frequency stimulation of the subthalamic nucleus suppresses oscillatory beta activity in patients with Parkinson’s disease in parallel with improvement in motor performance. J. Neurosci. 28, 6165–6173. 10.1523/JNEUROSCI.0282-08.2008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Kühn AA, Kupsch A, Schneider G-H, Brown P, 2006. Reduction in subthalamic 8–35 Hz oscillatory activity correlates with clinical improvement in Parkinson’s disease: STN activity and motor improvement. European Journal of Neuroscience 23, 1956–1960. 10.1111/j.1460-9568.2006.04717.x [DOI] [PubMed] [Google Scholar]
  17. Little S, Beudel M, Zrinzo L, Foltynie T, Limousin P, Hariz M, Neal S, Cheeran B, Cagnan H, Gratwicke J, Aziz TZ, Pogosyan A, Brown P, 2016a. Bilateral adaptive deep brain stimulation is effective in Parkinson’s disease. Journal of Neurology, Neurosurgery & Psychiatry 87, 717–721. 10.1136/jnnp-2015-310972 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Little S, Pogosyan A, Neal S, Zavala B, Zrinzo L, Hariz M, Foltynie T, Limousin P, Ashkan K, FitzGerald J, Green AL, Aziz TZ, Brown P, 2013. Adaptive deep brain stimulation in advanced Parkinson disease. Annals of Neurology 74, 449–457. 10.1002/ana.23951 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Little S, Tripoliti E, Beudel M, Pogosyan A, Cagnan H, Herz D, Bestmann S, Aziz T, Cheeran B, Zrinzo L, Hariz M, Hyam J, Limousin P, Foltynie T, Brown P, 2016b. Adaptive deep brain stimulation for Parkinson’s disease demonstrates reduced speech side effects compared to conventional stimulation in the acute setting. Journal of Neurology, Neurosurgery & Psychiatry jnnp-2016–313518. 10.1136/jnnp-2016-313518 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Lofredi R, Tan H, Neumann W-J, Yeh C-H, Schneider G-H, Kühn AA, Brown P, 2019. Beta bursts during continuous movements accompany the velocity decrement in Parkinson’s disease patients. Neurobiol. Dis. 127, 462–471. 10.1016/j.nbd.2019.03.013 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Manning JR, Jacobs J, Fried I, Kahana MJ, 2009. Broadband shifts in local field potential power spectra are correlated with single-neuron spiking in humans. J. Neurosci. 29, 13613–13620. 10.1523/JNEUROSCI.2041-09.2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Meidahl AC, Moll CKE, van Wijk BCM, Gulberti A, Tinkhauser G, Westphal M, Engel AK, Hamel W, Brown P, Sharott A, 2019. Synchronised spiking activity underlies phase amplitude coupling in the subthalamic nucleus of Parkinson’s disease patients. Neurobiol. Dis. 127, 101–113. 10.1016/j.nbd.2019.02.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Murthy VN, Fetz EE, 1996. Synchronization of neurons during local field potential oscillations in sensorimotor cortex of awake monkeys. J. Neurophysiol. 76, 3968–3982. [DOI] [PubMed] [Google Scholar]
  24. Murthy VN, Fetz EE, 1992. Coherent 25- to 35-Hz oscillations in the sensorimotor cortex of awake behaving monkeys. Proc. Natl. Acad. Sci. U.S.A. 89, 5670–5674. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Piña-Fuentes D, Beudel M, Little S, Brown P, Oterdoom DLM, van Dijk JMC, 2019. Adaptive deep brain stimulation as advanced Parkinson’s disease treatment (ADAPT study): protocol for a pseudo-randomised clinical study. BMJ Open 9, e029652 10.1136/bmjopen-2019-029652 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Piña-Fuentes D, Little S, Oterdoom M, Neal S, Pogosyan A, Tijssen MAJ, van Laar T, Brown P, van Dijk JMC, Beudel M, 2017. Adaptive DBS in a Parkinson’s patient with chronically implanted DBS: A proof of principle. Mov. Disord. 32, 1253–1254. 10.1002/mds.26959 [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Quinn EJ, Blumenfeld Z, Velisar A, Koop MM, Shreve LA, Trager MH, Hill BC, Kilbane C, Henderson JM, Brontë-Stewart H, 2015. Beta oscillations in freely moving Parkinson’s subjects are attenuated during deep brain stimulation: Beta Oscillations In Free Moving PD Subjects. Movement Disorders 30, 1750–1758. 10.1002/mds.26376 [DOI] [PubMed] [Google Scholar]
  28. Rosa M, Arlotti M, Ardolino G, Cogiamanian F, Marceglia S, Di Fonzo A, Cortese F, Rampini PM, Priori A, 2015. Adaptive deep brain stimulation in a freely moving parkinsonian patient: Letters: New Observations. Movement Disorders 30, 1003–1005. 10.1002/mds.26241 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Rosa M, Arlotti M, Marceglia S, Cogiamanian F, Ardolino G, Fonzo AD, Lopiano L, Scelzo E, Merola A, Locatelli M, Rampini PM, Priori A, 2017. Adaptive deep brain stimulation controls levodopa-induced side effects in Parkinsonian patients. Mov. Disord. 32, 628–629. 10.1002/mds.26953 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Schmidt SL, Peters JJ, Turner DA, Grill WM, 2020. Continuous deep brain stimulation of the subthalamic nucleus may not modulate beta bursts in patients with Parkinson’s disease. Brain Stimul 13, 433–443. 10.1016/j.brs.2019.12.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Shreve LA, Velisar A, Malekmohammadi M, Koop MM, Trager M, Quinn EJ, Hill BC, Blumenfeld Z, Kilbane C, Mantovani A, Henderson JM, Brontë-Stewart H, 2017. Subthalamic oscillations and phase amplitude coupling are greater in the more affected hemisphere in Parkinson’s disease. Clin Neurophysiol 128, 128–137. 10.1016/j.clinph.2016.10.095 [DOI] [PubMed] [Google Scholar]
  32. Syrkin-Nikolau J, Koop MM, Prieto T, Anidi C, Afzal MF, Velisar A, Blumenfeld Z, Martin T, Trager M, Bronte-Stewart H, 2017. Subthalamic neural entropy is a feature of freezing of gait in freely moving people with Parkinson’s disease. Neurobiol. Dis. 108, 288–297. 10.1016/j.nbd.2017.09.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Tinkhauser G, Pogosyan A, Little S, Beudel M, Herz DM, Tan H, Brown P, 2017a. The modulatory effect of adaptive deep brain stimulation on beta bursts in Parkinson’s disease. Brain 140, 1053–1067. 10.1093/brain/awx010 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Tinkhauser G, Pogosyan A, Tan H, Herz DM, Kühn AA, Brown P, 2017b. Beta burst dynamics in Parkinson’s disease OFF and ON dopaminergic medication. Brain 10.1093/brain/awx252 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Tinkhauser G, Torrecillos F, Duclos Y, Tan H, Pogosyan A, Fischer P, Carron R, Welter M-L, Karachi C, Vandenberghe W, Nuttin B, Witjas T, Régis J, Azulay J-P, Eusebio A, Brown P, 2018. Beta burst coupling across the motor circuit in Parkinson’s disease. Neurobiol. Dis. 117, 217–225. 10.1016/j.nbd.2018.06.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Torrecillos F, Tinkhauser G, Fischer P, Green AL, Aziz TZ, Foltynie T, Limousin P, Zrinzo L, Ashkan K, Brown P, Tan H, 2018. Modulation of Beta Bursts in the Subthalamic Nucleus Predicts Motor Performance. J. Neurosci. 38, 8905–8917. 10.1523/JNEUROSCI.1314-18.2018 [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Trager MH, Koop MM, Velisar A, Blumenfeld Z, Nikolau JS, Quinn EJ, Martin T, Bronte-Stewart H, 2016. Subthalamic beta oscillations are attenuated after withdrawal of chronic high frequency neurostimulation in Parkinson’s disease. Neurobiology of Disease 96, 22–30. 10.1016/j.nbd.2016.08.003 [DOI] [PubMed] [Google Scholar]
  38. Velisar A, Syrkin-Nikolau J, Blumenfeld Z, Trager MH, Afzal MF, Prabhakar V, Bronte-Stewart H, 2019. Dual threshold neural closed loop deep brain stimulation in Parkinson disease patients. Brain Stimul 12, 868–876. 10.1016/j.brs.2019.02.020 [DOI] [PubMed] [Google Scholar]
  39. Welch P, 1967. The use of fast Fourier transform for the estimation of power spectra: A method based on time averaging over short, modified periodograms. IEEE Transactions on Audio and Electroacoustics 15, 70–73. 10.1109/TAU.1967.1161901 [DOI] [Google Scholar]
  40. Whitmer D, de Solages C, Hill B, Yu H, Henderson JM, Bronte-Stewart H, 2012. High frequency deep brain stimulation attenuates subthalamic and cortical rhythms in Parkinson’s disease. Front Hum Neurosci 6, 155 10.3389/fnhum.2012.00155 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1609153-supplement-1.docx (442.9KB, docx)

RESOURCES