Temporal Dynamics of Microlesion Effects in Subthalamic Local Field Potentials

Abstract

We investigated how subthalamic local field potentials evolve as the microlesion effect emerges and wanes after electrode implantation in Parkinson's disease. Thirteen patients underwent repeated resting recordings that were analyzed across six predefined postoperative periods (days 0-6, 7-30, 31-90, 91-180, 181-365, and ≥366). Power spectral density (1-50 Hz) was decomposed into periodic and aperiodic components. Period-wise changes were tested with nonparametric within-subject analyses, and spatial differences across sensing-electrode pairs were evaluated with population-averaged regression under multiplicity control. Total local field potential power and aperiodic parameters (offset and exponent) followed an inverted-U trajectory, peaking at 31-90 days and declining by ≥12 months. In contrast, periodic beta power (13-30 Hz) increased from approximately 1-3 months onward and remained elevated at 6-12 months, resulting in a higher periodic-to-total beta ratio in late windows. Spatially, periodic beta was maximal over more dorsal, putative sensorimotor territories, whereas the aperiodic exponent was relatively larger ventrally, indicating distinct topographies of oscillatory versus aperiodic activity. Clinically, Movement Disorder Society-Unified Parkinson's Disease Rating Scale Part III improved at 6 months with partial attenuation by 12 months; time-matched correlations with electrophysiological metrics did not survive multiple-comparison adjustment. These findings suggest that the microlesion initially suppresses oscillatory beta more than broadband activity, with a later relative prominence of the periodic component, and that spatial dissociation between periodic and aperiodic features may inform biomarker selection and contact targeting for adaptive stimulation.

Introduction

Deep brain stimulation (DBS) has emerged as a highly effective therapeutic intervention for Parkinson's disease (PD).^1-3) By delivering electrical stimulation to specific brain regions, DBS modulates neural activity and alleviates motor symptoms. One key approach to understanding the mechanisms underlying the therapeutic effects of DBS is the analysis of the local field potentials (LFPs), which provides insight into treatment-related neural dynamics. LFPs have been extensively studied since the advent of DBS.^4-6)

A phenomenon known as the microlesion effect was observed in the months following DBS electrode implantation.^7-9) This effect refers to the temporary improvement in motor symptoms before the full therapeutic benefits of stimulation become evident and has been reported as a predictor of the short-term clinical effects of DBS in PD.⁹⁾ It is thought to arise from a combination of electrode-related tissue perturbation, inflammation, and functional alterations in the stimulated networks.^10,11) Characterizing how LFP signals change as the microlesion effect emerges and then wanes is therefore important for understanding postoperative physiology and optimizing DBS.

In recent years, LFP analyses have increasingly separated the spectrum into periodic (narrowband, oscillatory) and aperiodic (broadband, 1/f-like) components.^5,12,13) The periodic component encompasses rhythmic oscillatory activity within specific frequency bands, such as alpha, beta, and gamma oscillations. These oscillations are known to play critical roles in information processing, the coordination of neural networks, and motor control. Beta oscillations play a significant role in PD pathophysiology, and their modulation using DBS demonstrates promising therapeutic effects.¹⁴⁾ In contrast, the aperiodic component encompasses non-oscillatory broadband activity that eludes description by a single or narrow frequency band, and can be parameterized by an exponent (spectral slope) and offset. Prior work links the exponent to excitation-inhibition balance and its sensitivity to therapy, while the offset relates to overall broadband power and correlates with population spiking and hemodynamic signals.^12,15,16) This periodic/aperiodic decomposition provides complementary biological information to traditional band power measures.¹²⁾

However, it remains unclear how the periodic beta and aperiodic components of subthalamic nucleus (STN)-LFP evolve across the first postoperative year as the microlesion effect attenuates, and whether their spatial distributions within the STN differ. To address this gap, we analyzed longitudinal STN-LFP recordings from patients with PD undergoing DBS, decomposed spectra into periodic and aperiodic components, and tested temporal changes across six predefined postoperative periods. We further compared spatial topographies across sensing-electrode (SE) pairs and summarized motor outcomes at 6 and 12 months to relate signal trajectories to clinical course.

Materials and Methods

This study included patients with PD who underwent STN-DBS surgery at Osaka University between November 2020 and April 2022, had LFP recorded for at least 12 months after electrode implantation, and had at least 100 LFP recordings. The patients were diagnosed with PD based on the Movement Disorder Society Clinical Diagnostic Criteria for PD¹⁷⁾ and treated with levodopa. The study protocol was approved by the Ethics Committee of Osaka University (No. 20213). The study was conducted in accordance with the Declaration of Helsinki; written informed consent was obtained from all patients, and their rights to voluntarism and withdrawal were ensured. All patients received DBS electrodes and internal pulse generators (IPG) on the same day. In total, 11 patients received simultaneous bilateral electrode implantation, while two underwent bilateral implantations with the second side implanted after an interval of approximately 3 months. The electrode implantation procedure was performed under general anesthesia,¹⁸⁾ and microelectrode recordings were employed to assess neural activity in the STN. The distal (tip) contact of each lead was precisely positioned along the ventral border of the STN adjacent to the substantia nigra. In eight cases, a 4-contact electrode lead (model 3389; Medtronic Inc., Minneapolis, MN, USA) was bilaterally inserted into the STN, and in the remaining five cases, an 8-contact directional lead (model B33005; Medtronic Inc.) was bilaterally inserted and connected to an IPG (Percept PC, Medtronic Inc.). For cylindrical leads, contacts were numbered 0, 1, 2, and 3 from ventral to dorsal. In our implantation approach, contact 0 was positioned near the ventral STN border; contact 1, near the center; contact 2, near the dorsal border; and contact 3 just dorsal to the STN. Stimulation commenced approximately 9.7 days post-electrode implantation and persisted with clinically optimized continuous stimulation (Table 1). PD motor symptoms were assessed using the Movement Disorder Society-Unified Parkinson's Disease Rating Scale (MDS-UPDRS) Part III within 3 months before DBS surgery and at 6 and 12 months after surgery, with stimulation ON and medication ON. Levodopa equivalent daily doses (LEDD) were calculated 3 months before and 6 months after surgery.¹⁹⁾ The optimal stimulating contact was defined as the contact that proved most effective during monopolar review or clinical use within the first or second postoperative month.

Table 1.

Patient Characteristics

ID	Sex	Age at implantation (y)	Age at diagnosis (y)	Preoperative MDS-UPDRS					LEDD (mg)		6-mo MDS-UPDRS III (ON/ON)	Optimal contact
				I	II	III ON	III OFF	IV	Pre-op	6-mo		Left	Right
1	M	71	61	16	21	26	42	5	400	19	500	1	1
2	F	61	46	17	33	43	65	18	1310	—	930	1	2
3	F	46	37	9	20	34	56	15	1185	24	835	2	1
4	F	55	40	16	26	28	52	15	710	31	580	1	1
5	M	62	54	24	17	17	58	14	1600	27	915	1	2
6	F	71	62	20	37	17	59	17	1080	30	635	2	1
7	F	63	46	15	22	26	41	13	922.5	13	387.5	1	2
8	F	61	42	6	15	14	30	13	1025	11	850	1	2
9	M	69	48	24	46	23	71	11	720	19	440	1	2
10	F	56	26	6	14	26	62	11	800	16	450	1	1
11	F	70	55	23	27	34	65	15	450	16	350	2	2
12	F	70	59	14	14	15	36	9	1330	13	695	2	2
13	M	54	44	14	28	22	71	18	1075	16	880	2	2

“—” indicates not available. Preoperative III ON/OFF indicate medication state at baseline; 6-month values are stimulation ON, medication ON.

F: female; LEDD: levodopa equivalent daily dose; M: male; MDS-UPDRS: Movement Disorder Society-Unified Parkinson’s Disease Rating Scale

Clinical outcomes

MDS-UPDRS Part III (motor subsection) was collected preoperatively (≤3 months) and postoperatively at 6 and 12 months. When multiple assessments occurred within a window near a target time point, the assessment closest to 6 or 12 months was used. These clinical evaluations coincide with the late postoperative windows used in the period-based analyses described below.

Postoperative time windows (period definitions)

Postoperative time was discretized into six named periods relative to the date of lead implantation: period 1, days 0-6; period 2, days 7-30; period 3, days 31-90; period 4, days 91-180; period 5, days 181-365; and period 6, ≥366 days. For staged bilateral implantation, postoperative time was referenced to the implantation date of the analyzed hemisphere. Hemisphere-level metrics were aligned to each hemisphere's own postoperative time and averaged to the participant level where specified. Clinical assessments at 6 and 12 months correspond to periods 5 and 6, respectively.

LFP-recording procedure

LFP recordings were acquired during routine visits across the postoperative year, with data retrieval facilitated later using a communication device. The timing of LFP recordings was independent of the patient's last medication intake, and patients could be in either the “on” or “off” state. Recordings were obtained in a resting state with the eyes open, either while lying in bed or sitting comfortably in a chair, employing bipolar recordings for approximately 20 seconds between each electrode contact. Six LFP datasets were recorded using a bipolar configuration employing the four contacts available on the electrode and creating six unique pairs of contact combinations. Electrical stimulation persisted until immediately before the LFP-recording, ceased during the recording, and resumed immediately afterward. In this study, only ring-mode recordings were used; directional-mode recordings were excluded. The impedance of each electrode was measured during each LFP-recording session.

Preprocessing

Approximately 20-s LFP epochs were transformed into the frequency domain using a Fast Fourier Transform. For each bipolar recording, the power spectral density (PSD) was estimated over 1-50 Hz and partitioned into canonical bands: delta (1-4 Hz), theta (4-8 Hz), alpha (8-13 Hz), beta (13-30 Hz), and gamma (30-50 Hz).

Band-limited power SP_B was defined as the area under the PSD within band B

Equivalently,

where B denotes the target band and Δf is the frequency-bin width. “Total Power” corresponds to B = 1-50 Hz.

Recordings were analyzed according to the six postoperative periods defined above (period 1-6). Possible ECG or motion artifacts identified by the device's proprietary classifiers were flagged at export and excluded from analysis when present. The FOOOF (Fitting Oscillations & One-Over-F) algorithm, an open-source toolbox developed by Donoghue et al.,²⁰⁾ was used to decompose the original PSD into periodic and aperiodic components. The parameters of the aperiodic component, specifically the aperiodic offsets and exponents, were then computed. The FOOOF algorithm settings were configured with peak width limits set to 0.5-12, an infinite maximum number of peaks, a minimum peak height of 0.05, a peak threshold of 2.0, and a fixed aperiodic mode.

All metrics―total LFP power (1-50 Hz), aperiodic offset, aperiodic exponent, beta power (13-30 Hz), and periodic beta power (beta-range peak power after removing the aperiodic background)―were z-scored within-participant across all available recordings (mean = 0, standard deviation [SD] = 1) prior to modeling.

Statistical analysis of period-wise changes

For each LFP metric―total power (1-50 Hz), aperiodic offset, aperiodic exponent, beta power (13-30 Hz), and periodic beta power―z-scored session values were summarized to one value per-subject per-period by averaging that subject's available sessions within the period (after artifact exclusion). Thus, each subject contributed at most one observation per period. The primary analysis tested for within-subject period effects (periods 1-6) with the Friedman test, using complete cases only (subjects with non-missing values for all six periods for the metric under test). When the Friedman test was significant at α = 0.05, post hoc Wilcoxon signed-rank tests (two-sided; exact p-values; zero differences excluded) were performed for two prespecified families of contrasts:

1. Baseline contrasts: period 1 vs periods 2-6 (reported without multiplicity correction as planned contrasts);

2. Adjacent contrasts (2 vs 3, 3 vs 4, 4 vs 5, 5 vs 6), with Holm step-down adjustment within this family.

Figures display subject-level distributions (box-plus-spaghetti plots) with p-values for the planned contrasts. Analyses were performed in JMP Pro 17.1.0, with selected checks in Python (pandas/SciPy/statsmodels).

Data transformation for spatial comparisons

For spatial comparisons, periodic beta power was expressed in decibels (dB) to place subjects and electrode configurations on a common scale; the aperiodic exponent is unitless and was analyzed as is. Only ring-mode recordings were analyzed; directional-mode recordings were excluded. SE pairs were 0-1, 1-2, 2-3, 0-2, 1-3, and 0-3. Within each participant×SE pair, repeated sessions were summarized by the median, so each participant contributed at most one value per SE pair. Postoperative period was not included in this spatial analysis. We compared SE pairs using population-averaged regression that accounts for repeated measures within participants (generalized estimating equations [GEEs], Gaussian/identity; exchangeable working correlation; robust sandwich SEs), using 0-1 as the reference. We report estimated marginal means with 95% confidence intervals for each SE pair and Holm-adjusted p-values for all pairwise contrasts (family-wise α = 0.05). Missing observations were not imputed.

Spatial distribution and interindividual variability in the strength of periodic high- and periodic low-beta power

Low- and high-beta bands are thought to reflect different neural activities in the beta power domain. We conducted the following analyses to delineate the characteristics of periodic high- and low-beta powers expressed in decibels (dB). First, to elucidate the spatial distribution, we calculated periodic high-beta power and periodic low-beta power (in dB) for each recorded SE pair (0-1, 1-2, 2-3, 0-2, 1-3, 0-3), pooling artifact-free epochs from all participants and sessions across the entire observation period. Additionally, to uncover individual differences, we summarized for each participant the distributions of periodic high- and low-beta powers (in dB) across all SE pairs and sessions. These descriptive results are shown as box-and-whisker plots (median, interquartile range [IQR]; whiskers 1.5 × IQR; outliers not displayed), and postoperative period was not included in this analysis. Formal statistical comparisons of SE pairs are reported elsewhere in this manuscript.

Clinicophysiological correlation and visualization

To address clinical relevance, we computed time-matched Spearman's rank correlations between the change in MDS-UPDRS Part III (preoperative to follow-up; stimulation ON/medication ON) and two prespecified LFP metrics (periodic beta, aperiodic exponent) at 6 months (period 5) and 12 months (period 6). At each time point, p-values were adjusted for the two tests using Holm's step-down procedure (family-wise α = 0.05). For visualization, individual Part III trajectories (preoperative, 6-month, 12-month; ON/ON) were plotted with thin lines, and group means ± SD were overlaid; only participants with ≥2 time points were displayed.

Results

Participants

Thirteen patients were included (mean age at implantation 62.2 ± 7.8 years; disease duration 14.5 ± 6.2 years). After excluding device-flagged artifacts, an average of 891.5 ± 347.5 (mean ± SD) STN-LFP recordings per participant were available (range 294-1,292). Across postoperative periods, the number of recordings was: period 1, 1,738; period 2, 4,438; period 3, 488; period 4, 1,485; period 5, 1,689; period 6, 1,751. Preoperative MDS-UPDRS scores were as follows: Part I, 15.7 ± 6.1; Part II, 24.6 ± 9.6; Part III on, 25.0 ± 8.5; Part III off, 54.5 ± 13.3; Part IV, 13.4 ± 3.7. At 6 and 12 months postoperatively (stimulation ON, medication ON), MDS-UPDRS Part III scores were 18.4 ± 6.0 (n = 12; Subject 2 missing) and 21.3 ± 6.3 (n = 12; Subject 8 missing), respectively. Preoperative LEDD was 969.8 ± 350.1 mg, and at 6 months postoperatively, it was 649.9 ± 213.8 mg (Table 1). The optimal stimulus contact for the left STN, resulting in the most significant clinical improvement of Parkinsonian motor symptoms, was identified as “Contact 1” in eight patients and “Contact 2” in five patients. For the right STN, the optimal contacts were “Contact 1” in five patients and “Contact 2” in eight patients (Table 1).

Impedance

Mean impedance values for the 3,389-type electrode were 1,281, 1,733, 1,745, 1,775, 1,681, and 1,745 ohms at months 1-6 after implantation. For the B33005 lead, the corresponding means were 1,188, 2,227, 2,167, 2,197, 2,111, and 2,127 ohms. In both lead types, impedance increased during the first month and then stabilized thereafter.

Temporal dynamics of LFPs throughout the recording period

We observed a marked escalation in z-scored total LFP power, with significant increase compared with period 1 in periods 3 (p = 0.002), 4 (p = 0.002), 5 (p = 0.016), and 6 (p = 0.042); the adjacent-period analysis further showed an increase from period 2 to 3 (Holm-adjusted p = 0.004) (Figure 1A and Table 2). Across periods, total LFP power followed an inverted-U trajectory, peaking at period 3 and attenuating thereafter; the decrease from period 5 to period 6 showed a trend (Holm-adjusted p = 0.081; unadjusted p = 0.027).

Fig. 1.

Period-wise changes in LFP metrics (complete-case analysis).

A, Total LFP power (1-50 Hz, z-score). B, Aperiodic exponent (z-score). C, Aperiodic offset (z-score). For each metric, values were first averaged within-subject for each postoperative period and then plotted as per-subject distributions (box: median and IQR; whiskers: 1.5 × IQR). Gray “spaghetti” lines connect the same participant across periods. Asterisks mark p < 0.05 for baseline contrasts (period 1 vs periods 2-6; Wilcoxon signed-rank, two-sided, exact, unadjusted). Brackets with p-values denote Holm-adjusted results for adjacent contrasts (2-3, 3-4, 4-5, 5-6). Periods are postoperative days: 0-6, 7-30, 31-90, 91-180, 181-365, and ≥366.

IQR: interquartile range; LFP: local field potential

Table 2.

Period-Wise Changes in LFP Metrics: Friedman Test and Adjacent Wilcoxon Contrasts

Metrics	Friedman χ2	df	n (complete)	p (Friedman)	Adjacent contrast	n (paired)	W (exact)	p (Holm-adjusted)	Decision
Total LFP power (1-50 Hz)	27.90	5	12	<0.001	2 vs 3	12	1	0.004	Significance
					3 vs 4	12	38	0.970	n.s.
					4 vs 5	12	14	0.104	n.s.
					5 vs 6	12	11	0.081	n.s.
Aperiodic offset	33.12	5	12	<0.001	2 vs 3	12	0	0.002	Significance
					3 vs 4	12	25	0.301	n.s.
					4 vs 5	12	15	0.128	n.s.
					5 vs 6	12	5	0.015	Significance
Aperiodic exponent	32.38	5	12	<0.001	2 vs 3	12	0	0.002	Significance
					3 vs 4	12	22	0.303	n.s.
					4 vs 5	12	20	0.303	n.s.
					5 vs 6	12	5	0.015	Significance
Beta power (13-30 Hz)	20.48	5	12	0.001	2 vs 3	12	5	0.020	Significance
					3 vs 4	12	28	1.00	n.s.
					4 vs 5	12	26	1.00	n.s.
					5 vs 6	12	31	1.00	n.s.
Periodic beta power	21.14	5	12	<0.001	2 vs 3	12	5	0.020	Significance
					3 vs 4	12	38	1.00	n.s.
					4 vs 5	12	37	1.00	n.s.
					5 vs 6	12	37	1.00	n.s.
Periodic low beta (13-20 Hz)	18.48	5	12	0.002	2 vs 3	12	20	0.601	n.s.
					3 vs 4	12	33	0.678	n.s.
					4 vs 5	12	22	0.605	n.s.
					5 vs 6	12	20	0.605	n.s.
Periodic high beta (20-30 Hz)	18.81	5	12	0.002	2 vs 3	12	6	0.027	Significance
					3 vs 4	12	30	1.00	n.s.
					4 vs 5	12	21	0.529	n.s.
					5 vs 6	12	35	1.00	n.s.
Periodic beta ratio (exploratory)	20.23	5	12	0.001	2 vs 3	12	15	0.256	n.s.
					3 vs 4	12	39	1.00	n.s.
					4 vs 5	12	38	1.00	n.s.
					5 vs 6	12	21	0.529	n.s.

Notes: Values are within-subject z-scores summarized to one value per subject per period. Period effect was tested with the Friedman test on complete cases (n shown). Post-hoc Wilcoxon signed-rank tests were two-sided with exact p-values: baseline contrasts (P1 vs P2-P6) were pre-specified and reported unadjusted; adjacent contrasts (2-3, 3-4, 4-5, 5-6) were Holm-adjusted to control the family-wise α = 0.05. p < 0.001 are shown as “<0.001.”

LFP: local field potential; n: paired sample size; n.s.: not significant; P: period; W: exact Wilcoxon statistic

Moreover, the aperiodic offset and aperiodic exponent exhibited a similar trend, showing significant increases in periods 3 (p = 0.002 and p = 0.003), 4 (p = 0.002 and p = 0.005), and 5 (p = 0.021 and p = 0.021) compared with period 1. In the adjacent-period analysis, both metrics increase from period 2 to period 3 (p = 0.002 for each), and declined from period 5 to period 6 (Holm-adjusted p ≤ 0.015), indicating a period-3 peak followed by a late decline (Figure 1B-C and Table 2).

The dynamics within the total beta power mirrored those observed in the total LFP power, with significant increase compared with period 1 in periods 3 (p = 0.002), 4 (p = 0.003), 5 (p = 0.010), and 6 (p = 0.021); the adjacent-period analysis further showed an increase from period 2 to 3 (Holm-adjusted p = 0.020) (Figure 2A and Table 2). The periodic beta power also showed significant increase in periods 3 (p = 0.012), 4 (p = 0.007), 5 (p = 0.012), and 6 (p = 0.016) compared with period 1; the adjacent-period analysis further showed an increase from period 2 to 3 (Holm-adjusted p = 0.020) (Figure 2B and Table 2). In contrast to total beta, periodic beta did not show clear attenuation after period 3 and remained elevated through periods 5-6. The periodic beta ratio also showed significant increase in periods 3 (p = 0.012), 4 (p = 0.007), 5 (p = 0.005), and 6 (p = 0.005) compared with period 1, indicating a greater relative contribution of the periodic component from period 3 onward (Figure 2C and Table 2).

Fig. 2.

Period-wise changes in beta-related metrics (complete-case analysis).

(A) Total Beta Power, (B) Total Periodic Beta Power, and (C) Periodic Beta Ratio. Each panel shows boxplots (boxes = median and IQR; whiskers = 1.5 × IQR; outliers = small open circles) overlaid with gray “spaghetti” lines connecting within-participant means across periods. Values are z-scored within participant (per metric). Analyses were restricted to complete cases (participants observed at all six periods). Overall differences were tested with the Friedman test, and post hoc tests were performed only when the overall test was significant. Baseline comparisons (period 1 vs periods 2-6) used two-sided Wilcoxon signed-rank tests without multiplicity adjustment; p < 0.05 is marked with asterisks (*) above the corresponding box. Adjacent-period contrasts (2-3, 3-4, 4-5, 5-6) are shown with brackets and Holm-adjusted p-values. Y-axes display z-scores with matched scales across panels.

IQR: interquartile range

Spatial distribution of periodic beta power and aperiodic component of LFP in STN

We tested whether the spatial topography differed between the oscillatory and aperiodic components across SE pairs. Periodic beta power (dB) was highest for the adjacent pair 1-2, exceeding 0-1 and 2-3 (Holm-adjusted p < 0.05). Compared with the non-adjacent pairs 0-2 and 0-3, 1-2 did not differ significantly, whereas it differed from 1 to 3 (Holm-adjusted p < 0.05). These spatial patterns were consistent across the six postoperative periods (periods 1-6), with the rank ordering of SE pairs essentially unchanged for both metrics (Supplementary Figure S1). In contrast, the aperiodic exponent was largest among adjacent pairs for 0-1, yet 0-1 was significantly lower than all non-adjacent pairs (0-2, 1-3, 0-3) after Holm correction (all p < 0.05). These results indicate distinct spatial topographies for periodic beta and the aperiodic component across contact pairs (Figure 3; Supplementary Table S1).

Fig. 3.

Spatial distribution of periodic beta power and aperiodic exponent across sensing-electrode pairs.

A, Periodic beta power (13-30 Hz; periodic component after aperiodic removal) expressed in decibels (dB). B, Aperiodic exponent (unitless). Only ring-mode recordings were analyzed. For each subject and SE pair (0-1, 1-2, 2-3, 0-2, 1-3, 0-3), repeated sessions were summarized by the median so that each subject contributed at most one value per pair. Boxes show median and interquartile range; whiskers, 1.5 × IQR; outliers not displayed. Data are pooled across postoperative periods; period-wise patterns are shown in Supplementary Fig. S1. Group differences among SE pairs were evaluated with population-averaged regression (GEE; Gaussian/identity, exchangeable working correlation, robust SEs) with Holm correction for all pairwise contrasts; detailed statistics are provided in Supplementary Table S1 (n = 13 participants).

GEE: generalized estimating equation; IQR: interquartile range; SE: sensing-electrode

Exploratory comparison of periodic low- and high-beta

Periodic low-beta (13-20 Hz) and high-beta (20-30 Hz) powers (in dB) showed broadly similar spatial profiles across SE pairs, while between-individual differences in band dominance were evident (some participants showing relatively stronger low-beta, others high-beta). Because no consistent, group-level divergence emerged between the two sub-bands, the primary analyses focused on aggregate periodic beta (13-30 Hz). Detailed distributions are provided in Supplementary Figure S2.

Clinical outcomes

MDS-UPDRS Part III (stimulation ON, medication ON) improved from 25.0 ± 8.5 preoperatively to 18.4 ± 6.0 at 6 months (n = 12) and 21.3 ± 6.3 at 12 months (n = 12), indicating greater improvement at 6 months with partial attenuation by 12 months (Table 1; Supplementary Figure S3). No association survived Holm adjustment at either time point. At 12 months, the aperiodic exponent showed a negative trend with clinical improvement (Spearman's ρ = −0.518, unadjusted p = 0.084; Holm-adjusted p = 0.168).

Discussion

Our aim was to delineate how subthalamic LFPs evolve as the microlesion effect emerges and wanes after STN-DBS lead implantation. We observed that total LFP power and aperiodic parameters (offset and exponent) followed an inverted-U trajectory, peaking during 31-90 days after implantation (period 3) and declining by ≥12 months (period 6). In contrast, periodic beta power increased from period 3 onward and remained elevated through periods 5-6, yielding a higher periodic-to-total beta ratio in the late windows. These patterns indicate that, as the microlesion effect diminishes, the periodic (narrowband) beta component accounts for an increasing share of beta-band power, whereas the broadband/aperiodic component declines.

In PD, motor worsening has been linked to transient increases in beta oscillations (“beta bursts”); periodic beta after removal of the aperiodic background can be viewed as a threshold-free proxy of oscillatory beta that encompasses burst-like activity. Accordingly, our data are compatible with the notion that the microlesion initially suppresses oscillatory beta more strongly than broadband activity, and as its influence diminishes, the oscillatory (periodic) component becomes relatively more prominent. We did not compute threshold-defined bursts; instead, we modeled the periodic component to mitigate threshold bias.²¹⁾ Recent work links the aperiodic exponent to excitatory-inhibitory (E/I) balance― higher exponent indexing greater inhibition― and shows increases with levodopa or DBS and decreases with drug withdrawal in PD.^15,22) Placed in that framework, our exponent peak at 1-3 months is consistent with a transient inhibition-dominated network state plausibly induced by the lesion, followed by attenuation of inhibition toward one year. Although our time-matched correlations between LFP metrics and MDS-UPDRS Part III did not survive correction, the clinical trajectory―greater improvement at 6 months with partial attenuation at 12 months with stimulation ON and medication ON―does not contradict this physiological arc.

Short-term studies (hrs to ～1 month) under controlled medication states report increased beta at one month in the OFF state, with aperiodic components reduced at one month and partly restored by levodopa.²³⁾ Our dense sampling across the first postoperative year complements those findings: periodic beta remains relatively elevated at 6-12 months, while aperiodic metrics peak earlier and then declined. Differences across reports likely reflect time window, medication/stimulation state during recording, and the spectral metric under study.

A central methodological aspect of our work is the explicit separation of periodic and aperiodic spectral components (FOOOF).²⁰⁾ This distinction matters because threshold-based burst metrics have been reported to be sensitive to amplitude drifts and parameter choices, and may therefore show little change even when band-limited beta power decreases with DBS stimulation.²¹⁾ By modeling periodic beta after removing the aperiodic background, our longitudinal analysis reduces threshold bias and suggests that the relative contribution of oscillatory beta increases from ～1 to 3 months onward as the microlesion effect fades.

We also examined spatial organization. After aperiodic removal, periodic beta was maximal over more dorsal sensorimotor territories, whereas the aperiodic exponent was relatively larger ventrally, and these topographies were consistent across periods 1-6 (Supplementary Figure S1). This dissociation aligns with the notion that periodic beta indexes local oscillatory synchrony, while the aperiodic exponent tracks broader E/I balance and may be more susceptible to mixed sources and volume conduction; whitening approaches similarly unmask intrinsic high-beta generators within STN.²⁴⁾

How do these findings fit with long-term observations after DBS? Off-therapy, long-horizon recordings (3-5 years) found no progression of STN beta dynamics (alpha increased, beta stable).²⁵⁾ Rather than contradicting those data, our results appear temporally complementary: we highlight transient first-year dynamics―an inverted-U in total/aperiodic power and a sustained relative prominence of periodic beta―that may settle into the long-term stability described by Anderson et al.²⁵⁾ once the microlesion-related state resolves.

Limitations include heterogeneity in medication/stimulation states across sessions, the bipolar montage (which reduces but does not eliminate volume conduction), relatively short epochs (～20 s), and limited systematic clinical sampling in the 31-90-day window. Nonetheless, the large number of within-subject recordings, complete-case nonparametric testing across six predefined periods, and the explicit periodic/aperiodic decomposition provide a coherent view of microlesion-phase physiology and may inform biomarker choice for adaptive DBS (e.g., favoring periodic beta when the aperiodic background is labile). With larger longitudinal cohorts, this decomposition framework can test whether STN-LFP dynamics (periodic beta and aperiodic metrics) differ by age at onset, disease duration, sex,²⁶⁾ or genetic background, and whether these factors modify their links to motor benefit.

Disclaimer

Author Haruhiko Kishima is one of the Editorial Board members of the Journal.

This author was not involved in the peer-review or decision-making process for this paper.

Conflicts of Interest Disclosure

All authors have no conflict of interest.

Ethical Approval

The Ethics Committee of Osaka University (No. 20213) approved the study protocol. The study was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all patients, and their rights to voluntarism and withdrawal were ensured. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this work is consistent with those guidelines.

Data Statement

The data that support the findings of this study are available from the corresponding author, Naoki Tani, upon reasonable request.

Supplementary Material

Supplementary Figure 1

Period-wise spatial distribution of periodic beta and aperiodic components in the STN.

Periodic beta power (13-30 Hz; dB) (A) and aperiodic exponent (unitless) (B), shown by postoperative period (see Methods). Ring-mode recordings only; SE pairs: 0-1, 1-2, 2-3, 0-2, 1-3, 0-3. For each subject × SE pair × period, repeated sessions were summarized by the median and displayed as boxplots (median, IQR; whiskers 1.5 × IQR). Pooled (period-agnostic) summaries appear in Figure 3; pairwise SE pair statistics (GEE, Holm-adjusted) are in Supplementary Table S1. (n = 13; per-period n varies.)

GEE: generalized estimating equation; IQR: interquartile range; SE: sensing-electrode; STN: subthalamic nucleus.

Supplementary Figure 2

Periodic low- vs high-beta power (dB): spatial pattern and interindividual variability

A, Horizontal boxplots of periodic low-beta (13-20 Hz) and periodic high-beta (20-30 Hz) across sensing-electrode (SE) pairs (0-1, 1-2, 2-3, 0-2, 1-3, 0-3). Artifact-free, ring-mode recordings were pooled across all sessions/periods; boxes show median and IQR, whiskers 1.5 × IQR; outliers not shown.

B, Per-subject boxplots (pooled across SE pairs) illustrating interindividual differences in low- and high-beta power. No formal statistics are presented here; see main text for inferential analyses.

IQR: interquartile range.

Supplementary Figure 3

MDS-UPDRS Part III over time (ON stimulation/ON medication) Thin colored lines show individual patient trajectories; black squares with error bars indicate the group mean ± SD. Assessments were obtained preoperatively, at 6 months, and at 12 months after implantation. Sample sizes: n = 13 pre-op; n = 12 at 6 and 12 months (only participants with ≥2 time points shown).

SD: standard deviation.

Funding Statement

This study was funded by the Japan Science and Technology Agency under the JST FOREST program (Grant Number: JPMJFR205F) and by JSPS KAKENHI (Grant Numbers JP23K08519, Japan). The funder did not play any role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the article for publication.