This study was a single-center observational study, and the study protocol was approved by the Ethics Committee of Zhujiang Hospital of Southern Medical University (Ethics Approval No. 2021-KY-123). All patients participated in this study after obtaining informed consent and in accordance with the Declaration of Helsinki. Patients were included and excluded from this study according to the following criteria: Inclusion criteria: 1. Diagnosed with primary Parkinson’s disease according to the Chinese Diagnostic Criteria of Parkinson’s Disease (2016 edition) (Parkinson’s Disease and Movement Disorders Group, Chinese Society of Neurology, 2016); 2. Having the indications for DBS surgery as described in the Chinese Expert Consensus on Parkinson’s Disease Deep Brain Electrical Stimulation Therapy (Second edition) (Group of Functional Neurosurgery, Chinese Society of Neurosurgery, 2020); 3. Surgical selection of the bilateral STN nuclei; and Exclusion criteria: 1. Significant cognitive dysfunction (Mini-mental State Examination <24 points); 2. medical coexisting diseases affecting surgery or survival; 3. missing clinical assessment data or unsuccessful recording of STN electrophysiologic signals). The flow diagram depicting patient enrollment is illustrated in Figure 1. Ultimately, a total of 15 patients were included, and the analysis encompassed LFP signals derived from 30 lateral STN nuclei.

Preoperatively, all patients were scanned with a 3.0 T head magnetic resonance imaging (MRI) in axial and coronal positions based on the anterior-posterior commissure (AC-PC) planes, and the scanning sequences were T1-weighted images (layer thickness 1 mm, layer spacing 0 mm), T2-weighted images (layer thickness 2 mm, layer spacing 0 mm), and susceptibility weighted imaging. All anti-Parkinson’s disease medications were discontinued 12 h before implantation. On the day of implantation, a Leksell stereotactic head frame (Elekta Instrument AB, Stockholm, Sweden) was installed under local anesthesia, with the base of the head frame parallel to the plane where the AC-PC was located and with the midline located in the mid-sagittal plane. Cranial Computed tomography (CT) (layer thickness 0.625 mm) was performed after the head frame was installed. Image fusion of MRI and CT images was performed using the Leksell stereotactic system (Elekta Instrument AB, Stockholm, Sweden) to guide the design of STN target coordinates and the optimal trajectory for electrode placement. The bilateral STN-DBS procedure was performed under intravenous and inhalation combined general anesthesia. The location and depth of the STN were confirmed by single-channel microelectrode recording during the procedure, the optimal placement of the DBS leads was verified by intraoperative CT, Surgical complications were excluded, and IPG was performed subcutaneously immediately after confirming the exact lead position.

To confirm the placement of the DBS leads, we used Lead-DBS version 2.3.2 software (https://www.lead-dbs.org/) for postoperative reconstruction localization (Horn and Kühn, 2015; Horn et al., 2017). After correction for inhomogeneity, postoperative thin-layer CT and preoperative MRI images were registered using SPM12 software (https://www.fil.ion.ucl.ac.uk/spm/software/spm12/). Afterward, the postoperative images were non-linearly normalized to the Montreal Neurological Institute (MNI) template brain (International Consortium of Brain Mapping 2009b nonlinear asymmetric) (Avants et al., 2011; Ewert et al., 2019) using the “effective low variance” Advanced Normalization Tools and Symmetric image Normalization approach DBS leads were manually localized following initial, semiautomated trajectory reconstruction. The resultant lead models were warped to MNI space using the aforementioned transforms. The method is consistent with previously reported reconstruction methods (Horn et al., 2017; Darmani et al., 2022). Figure 2 delineates the bilateral STN-DBS lead locations for the entire subject cohort (n = 15).

LFP data were collected by an IPG equipped with wireless sensing technology (SR1101, SceneRay, Suzhou, China) (Mandali et al., 2021) postoperatively in the hyperacute period: recording the STN-LFP data of the Medication-OFF state at 6 h, 12 h, 24 h, 36 h, and 48 h postoperatively (Session1-5); One month postoperatively (before active stimulation): STN-LFP signals of Medication-OFF state (Session6) and Medication-ON state (Session7) were recorded.

Each DBS electrode has four contacts, the left electrode contacts are numbered as 0, 1, 2, 3 (where contact 0 position is the deepest and contact 3 position is the most superficial), and the right electrode contacts are numbered as 4, 5, 6, 7 (where contact 4 position is the deepest and contact 7 position is the most superficial). For postoperative LFP signals, LFP was obtained from adjacent contact pairs by wireless sensors and digitized by an analog-to-digital converter at a sampling rate of 1,000 Hz. Acquisition criteria: (1) Medication-OFF state at least medication stopping time ≥12 h, and Medication-ON state at least medication taking time ≥1 h; (2) Recording duration was 180 s, and the acquisition contact pairs were Left-STN (0&1, 1&2, 2&3) and Right-STN (4&5, 5&6, 6&7); and (3) The patient was in resting state, with eyes quietly closed and lying flat on the bed.

The LFP data were browsed after acquisition, and records with poor data quality were reacquired. We determined the relative positions of the electrode contacts and the STN by postoperative image fusion reconstruction, and the contact located within the motor sub-region of the STN were included in the analysis. Eventually, a total of 204 LFP spectra were included in the analysis. The relevant data are showed in Supplementary Table S1.

All data preprocessing steps and subsequent analyses were performed using MATLAB (Version 2020a, MathWorks, Inc., Natick, MA, USA) and Python (Version 3.7). First, the local field potential data were bandpass filtered using the pop_firws function in the EEGLAB toolbox (http://www.sccn.ucsd.edu/eeglab/) with lower and upper cutoff frequencies of 1 and 499 Hz, respectively, then apply notch filter at 50 Hz as well as its harmonics (Delorme and Makeig, 2004). The data were then manually inspected for periods of artefact, which were marked for exclusion. Power spectral density (PSD) estimation was performed using the Welch method with a 0.5-s Hamming window, 50% overlap, and a 0.5-Hz resolution. Then, the SpecParam Python package was used to separate periodic and aperiodic components from the spectral results (Donoghue et al., 2020b). The power spectral density P f for each frequency f is expressed as: P f = L f + ∑ n G n f . Where the P f is a combination of the aperiodic component L f and Gaussians G n f . The periodic components were parameterized as a mixture of Gaussian distributions: G n f = α ∗ exp − f − c 2 2 ∗ ω 2 . With α as the height of the peak, over and above the aperiodic component, c as the center frequency of the peak, ω as the width of the peak, and f as the array of frequency values. The aperiodic component was also parameterized using a Lorentzian function as: L f = b − ⁡ log   k + f χ . With b as the broadband “offset,” k as the “knee,” and χ as the ‘exponent’ of the aperiodic fit. The full model was described using these periodic and aperiodic parameters, the goodness of fit was estimated by comparing each fit with the original PSD in terms of the mean absolute error and the R 2 of the fit. To avoid overfitting, peak_width_limits were set at 1–12, the fitted frequency range was 1–38 Hz, and aperiodic_mode was chosen as “fixed” (fitted without a knee parameter) considering the narrow frequency range. For further analysis, all the models meet the condition that R 2 value is higher than 0.95. After completing the parameterization of the power spectra, the fitted spectra were used to subtract the aperiodic components to obtain periodic oscillatory components. It is worth noting that the result of the SpecParam algorithm is a logarithmic scale, so we first converted it to a linear scale before subtraction. At last, we use bandpower function in MATLAB to calculate the power values, which means calculating the average power in a frequency interval by integrating the power spectral density estimate. We investigated band-limited power of fitted spectra, aperiodic, and periodic components in three frequency bands traditionally used in EEG analysis: theta (4–8 Hz), alpha (8–13 Hz), and beta (13–30 Hz).

Paired t-test was used to study the change in motor symptoms in patients preoperatively and 1 month postoperatively under Medication-OFF state, and the effect of levodopa medication on the aperiodic component (exponents and offsets), periodic component (power values) and impedance at 1 month postoperatively. One-way repeated-measure analyses of variance (ANOVA) with Bonferroni’s multiple comparison test was applied to investigate whether the aperiodic components (exponents and offsets), periodic components (power values), and impedance varied with sessions. All statistical analyses were performed with SPSS 25.0 software (SPSS Inc., Chicago, Illinois, USA), and the significance level was set at α = 0.05.

Our results showed that there were within-group differences in raw theta power but not periodic theta power (raw_theta_power: F = 3.431, p < 0.01; periodic_theta_power: F = 1.295, p > 0.05, refer to Figures 4A, B). In the postoperative hyperacute period, there were no significant differences in both raw and periodic theta power, and in the postoperative 1 month, the state of the levodopa medication also had no significant effect on raw and periodic theta power. However, we found significant differences between the Medication-ON state in 1 month postoperatively and the raw theta power at 12 h postoperatively and 48 h postoperatively.

Results showed within-group differences in both raw and periodic alpha power (raw_alpha_power: F = 15.587, p < 0.01; periodic_alpha_power: F = 9.092, p < 0.01, refer to Figures 4C, D). There was no significant difference between raw and periodic alpha power in the postoperative hyperacute period. In the postoperative 1 month, the levodopa Medication-ON or Medication-OFF state had no significant effect on raw and periodic alpha power. Of note, there was a significant difference in both raw and periodic alpha power in the comparison of Medication-ON and Medication-OFF state in the postoperative 1 month and in the postoperative hyperacute period.

For the beta band, our results showed within-group differences in both raw and periodic beta power (raw_beta_power: F = 12.03, p < 0.05; periodic_beta_power: F = 5.85, p < 0.05, refer to Figures 5A, B). Multiple comparisons showed no significant differences in both raw and periodic beta power values during the postoperative hyperacute period, whereas significant differences existed between the postoperative hyperacute period and the postoperative 1 month Medication-OFF state. For comparisons between the postoperative hyperacute period and the Medication-ON state, differences were found in raw beta power but not in periodic beta power.

Of note, levodopa medication at 1 month postoperatively effectively reduced beta power values (raw_beta_power: t (29) = 4.78, p < 0.05, d = 0.87, 95% CI [0.26; 0.66]; periodic_beta_power: t (29) = 6.06, p < 0.05; d = 7.14, 95% CI [0.18; 0.35]), but the above significant differences did not survive correction for multiple comparisons in our study.

We compared changes in aperiodic components (exponents, offsets) over time over 7 sessions of visits, and the results showed within-group differences (exponents: F = 6.792, p < 0.05; offsets: F = 4.612, p < 0.05, refer to Figures 5C, D). Multiple comparison analyses showed no significant differences in either exponent or offset in the postoperative hyperacute period, nor in the comparison of 1 month postoperative Medication-ON state to the postoperative hyperacute period. However, we found a significant difference in the exponents and offsets between the postoperative 1 month Medication-OFF state and the postoperative hyperacute period.

Note that levodopa medication had an effect on aperiodic components at 1 month postoperatively (exponents: t (29) = −7.02, p < 0.05, d = −1.28, 95% CI [−0.3; −0.16]; offsets: t (29) = −5.12, p < 0.05; d = −0.9, 95% CI [−0.25; −0.1]). Again, the above significant differences did not survive correction for multiple comparisons in our study.

We analyzed impedance in the postoperative hyperacute period and in the postoperative 1 month period and showed that the impedance values in the postoperative 1 month period were significantly higher than in the postoperative hyperacute period and statistically significant differences existed (Impedance: t (29) = −7.4, p < 0.05, d = −1.3, 95% CI [−1.1; −0.63]).