Supplementary Materials Differential contributions of subthalamic beta rhythms and 1/f broadband activity to motor symptoms in Parkinson’s disease Stephanie Martina,b†, Iñaki Iturratea†*, Ricardo Chavarriagaa, Robert Leeba, Aleksander Sobolewskia, Andrew M. Lia,c, Julien Zaldivard,e, Iulia Peciu-Florianud, Etienne Pralongd, Mayte Castro-Jiménezd, David Benningerd, François Vingerhoetsd, Robert T. Knightb,f, Jocelyne Blochd†, José del R. Millána† † a
equal contribution Defitech Chair in Brain-Machine Interface, Center for Neuroprosthetics, Ecole Polytechnique Fédérale de
Lausanne, Switzerland b
Helen Wills Neuroscience Institute, University of California, Berkeley, CA, USA
c
Li Ka Shing Faculty of Medicine, University of Hong Kong, Hong Kong, P.R. China
d
Department of Clinical Neurosciences (Neurology and Neurosurgery), University Hospital of Vaud (CHUV),
Lausanne, Switzerland eService f
de Neurochirurgie, Hôpital de Sion, Hôpital du Valais, Switzerland.
Department of Psychology, University of California, Berkeley, CA, USA
*Corresponding author: Iñaki Iturrate, CNBI, Center for Neuroprosthetics, EPFL, Ch. des Mines 9, 1202 Genève, Switzerland,
[email protected] Acronyms LFP
Local field potentials
STN
Subthalamic nucleus
DBS
Deep brain stimulation
UPDRS Unified Parkinson's Disease rating scale CT
Computerized tomography
MRI
Magnetic resonance imaging
PSD
Power spectral density
FDR
False discovery rate
1
Supplementary methods Subjects and experimental protocol LFPs were recorded between 36 and 48 hours after STN DBS bilateral implantation in thirteen subjects using a 16-channel amplifier (gUsbAmp, g.tec, Austria) and sampled at 512 Hz. All patients volunteered and gave written informed consent. Experimental protocol was approved by the local ethical committee (CER-VD, Commission Cantonale (VD) d’Ethique de la recherche sur l’être humain). Clinical characteristics of each patient are presented in Supplementary Table 1. Patients sat comfortably in a semi-recumbent position in their hospital bed. LFPs were recorded bilaterally during six minutes in a resting state condition. During the recording, patients were asked to restrain from performing any movements nor speaking, and maintaining their eyes open. Prior to recording, the stimulation device was turned off for at least ten minutes. Patients were instructed to relax and keep their eyes open during the recording. Patients were clinically assessed at the end of the recording by independent neurologists using the Unified Parkinson's Disease Rating Scale part III Motor Examination (UPDRS III) sub-scores for bradykinesia (item 23), rigidity (item 22) and resting tremor symptoms (item 20) for both upper limbs.
Surgical procedure Long-acting dopaminergic medication was withdrawn 48 hours prior and short-acting medication was withdrawn 12-20 hours prior to off-medication pre-operative testing and to the DBS lead implantation procedure, always performed by the same surgeon. The awake surgical procedure stereotactically implanted a 3389 Medtronic electrode in the motor part of the STN. Direct surgical planning was performed with the Medtronic stealth station, based on a CT scan obtained with the CRW stereotactic frame, and fused with two 3 Tesla MRI sequences (MPRAGE with Gadolinium and space T2), obtained prior to surgery. The precise trajectory of the implanted electrode was simulated and its exact coordinates were calculated in relation with the position of the stereotactic frame. During the surgery, a cannula was implanted towards the STN following the calculation of the trajectory and remained in place until the end of the surgery. Through this cannula, a microelectrode was lowered toward the target in the STN and the border of the motor STN was visualized as the transition from slow spiking neurons to fast spiking neurons. After a first phase 2
of microrecording, macrostimulation was performed under neurological clinical assessment; improvement of rigidity was expected in case of correct placement. Adjustment of the electrode placement through a second track was eventually performed to optimize the clinical response to stimulation. The electrode location was checked during the surgery with the O-Arm (Medtronic). An externalized extension cable was connected to the distal part of the electrode and tunneled posterior to the skin incision to record the brain activity for a period of 3 days. Internalization of the electrode and connection of the electrode to an internal stimulator were performed at day 4 under general anesthesia. Once the definitive electrode was placed through the same cannula to the definitive target, a second CT scan was obtained after having implanted the definitive electrode. Finally, the electrode was fixed with cement on the skull. In order to assess the efficacy of stimulation, we calculated the clinical score during OFF- versus ON-stimulation. Results showed that the clinical scores were significantly better during ONstimulation than during OFF-stimulation across patients and hemispheres (meanON= 2.6±2.2, meanOFF= 5.2±3.9; p=0.0001, signed rank test). The improvement in clinical score from OFF- to ON-stimulation was calculated as follows:
𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖(%) = 100 ∙
𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑂𝑂𝑂𝑂𝑂𝑂 − 𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑂𝑂𝑂𝑂 𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑂𝑂𝑂𝑂𝑂𝑂
The average improvement across patients and hemisphere was 43±7%, similar to those obtained by previous works1.
Power spectrum modeling Six minutes of bipolar LFP activity were derived from two contacts, selected based on the electrodes chosen by the neurosurgeon for DBS stimulation. LFPs were first high-pass filtered at 1 Hz using a second order Butterworth filter. Artifact were removed using an automatic threshold approach (data with activity higher than 50µV in absolute value), followed by visual inspection and rejection of contaminated portions of data. Those windows with abnormal activations (higher than 50 µV for the automatic approach; or marked as noisy by visual inspection) were eliminated from further analyses. Altogether, the recordings were very clean in all participants, but one 3
(Patient 10). The amount of data rejected for this patient was almost 20%, whereas it was 0.1; Supplementary Table 3). These findings agree with previous electrophysiological studies showing that pathological beta activity was associated with bradykinesia and rigidity, but not tremor8. Similar to the results obtained with our model, the correlations between motor symptoms and beta amplitude using actual PSDs were significant only when removing the broadband activity (FDR corrected p0.05). In addition, the increase in correlation between with and without broadband activity was significant for Rigidity (Hotelling’s t-test, FDR corrected p=0.02) and B+R (FDR corrected p=0.03). We found no significant differences between the results using the model and those obtained using real data (FDR corrected p>0.2 for all motor symptoms). Despite this, the model approach (without broadband) still explained more variance (Bradykinesia=30%, Rigidity=47%, Tremor=0.45%, B+R=46%, B+R+T=26%) than the real data (without broadband; Bradykinesia=25%, Rigidity=45%, Tremor=1%, B+R=41%, B+R+T=20%). To further evaluate the validity of our model, we compared the beta amplitude and beta frequency estimated by the model with those obtained using the actual PSDs. Results showed a very strong correlation between the beta power in the real data and the modeled beta peak (Supplementary Figure 2; r=0.97; p=10-7). Similarly, the correlation between the modeled beta frequency and the real beta frequency was also very high (r=0.62, p=0.0006), further highlighting the accuracy of the model.
9
Supplementary figures
Supplementary Figure 1. Model fits to individual datasets. Model fits (orange = broadband activity, blue = beta activity and green = sum of broadband and beta activity) for each patient and hemisphere, together with the actual power spectral densities (black lines). For each subject and hemisphere, the bradykinesia, rigidity and tremor scores are also reported, as well as the individual R2 goodness of fit. Note that the fitting was not performed (nor shown) within the 48-52Hz due to the use of a notch filter.
Supplementary Figure 2. Comparison between the modeled and actual data for the beta amplitude (left panel) and the beta frequency (right panel) across patients and hemispheres.
10
Supplementary tables Supplementary Table 1: Clinical details of patients
1 2
Patient
Disease duration (years)
Age (years)
Gender (M/F)
P01
15
60
M
3
7
2
4
1
3
0
0
702
P02
11
81
F
4
4
2
2
0
0
4
0
950
P03
7
62
M
8
6
4
3
4
3
0
0
1143
P04
15
68
M
8
11
4
4
4
4
0
3
1250
P05
22
56
M
7
6
3
4
3
2
1
0
613
P06
10
70
M
6
7
4
3
1
2
1
2
1009
P07
11
77
M
1
1
1
0
0
1
0
0
1052
P08
14
65
F
7
2
3
1
0
0
4
1
1144
P09
16
74
M
5
4
3
2
2
1
0
1
911
P10
6
62
F
6
0
1
0
1
0
4
0
1512
P11
15
70
M
5
3
2
1
2
2
1
0
994
P12
18
67
M
2
0
0
0
2
0
0
0
1384
P13
11
66
M
3
4
1
2
2
2
0
0
1324
µ±SEM
13.2±4.4
67.5±6.8
N/A
5.0±2.3
4.2±3.2
2.3±1.3
2.0±1.5
1.7±1.4
1.5±1.3
1.2±1.7
0.5±1.0
1076.0±257.4
UPDRS III (23+22+20) left right
Bradykinesia (UPDRS III - 23) left right
Rigidity (UPDRS III - 22) left right
Tremor (UPDRS III - 20) left right
Drugs prior to surgery (LED mg)2
Sum of UPDRS III items 20, 22, 23 and both hemispheres, during OFF-medication and OFF-stimulation conditions. Levodopa Equivalent Dose.
11
Supplementary Table 2: Model parameters Parameter
Lower
Upper
Initial
bound
bound
value
𝒌𝒌𝟏𝟏 , offset
-10
10
0
0.01
20
10
𝒌𝒌𝟑𝟑 , power
0.01
2
1
𝒌𝒌𝟒𝟒 , amplitude
0
max(PSD)
max(PSD)
𝒌𝒌𝟓𝟓 , mean
8
35
21.5
𝒌𝒌𝟔𝟔 , standard deviation
2
7
4.5
𝒌𝒌𝟐𝟐 , rotational offset
Arbitrary units
Supplementary Table 3: Relationship between model parameters and motor symptoms Parameter
mean±std
corr(B)
corr(R)
corr(T)
corr(B+R)
corr(B+R+T)
𝒌𝒌𝟏𝟏 , offset
-4.33±2.93
-0.04
0.14
0.11
0.08
0.04
𝒌𝒌𝟐𝟐 , rotational offset
9.80±3.61
-0.25
0.01
-0.01
-0.15
-0.17
𝒌𝒌𝟑𝟑 , power
0.50±0.33
-0.08
0.07
0.08
0.02
0.04
𝒌𝒌𝟒𝟒 , amplitude
0.65±0.25
0.55*
0.69**
-0.07
0.68**
0.50*
26.49±4.58
-0.10
-0.21
-0.04
-0.17
-0.18
6.96±0.22
0.25
0.32
0.01
0.32
0.25
𝒌𝒌𝟓𝟓 , mean
𝒌𝒌𝟔𝟔 , standard deviation
*P