Subthalamic beta activity and neuroimaging concordance in deep brain stimulation: electrode placement and clinical outcomes Authors: Malen Razkin, Edurne Ruiz de Gopegui, Beatriz Tijero, Tamara Fernández del Valle, Gaizka Bilbao, Ainara Dolado, Imanol Lambarri, Juan Carlos Gómez-Estéban, Asier Erramuzpe, Marta Ruiz-Lopez Published: 2026 ABSTRACT AbstractObjectiveDeep brain stimulation (DBS) of the subthalamic nucleus (STN) is an established therapy for advanced Parkinson's disease (PD), particularly in managing levodopa-induced motor complications and refractory tremor. The STN is the preferred target for high-frequency DBS, where precise electrode placement is key to optimal outcomes. This study assesses the reliability of local field potential (LFP) recordings in identifying STN electrode placement by comparing them with imaging and atlas based reconstruction methods.MethodsLFP beta-band activity was recorded 15 days post-surgery in 59 PD patients with DBS implants. Electrode locations were classified as optimal, suboptimal, or outside the STN us-ing neuroimaging reconstructions. Clinical outcomes were assessed at one year using the Unified Parkinson's Disease Rating Scale (UPDRS I–IV).ResultsBeta frequency, but not magnitude, differed significantly across placement categories, with optimal electrodes showing higher beta frequency than suboptimal ones. Sensing-identified contacts showed strong concordance with neuroimaging-identified contacts across hemispheres. Op-timal electrode placements were associated with significantly greater improvements in UPDRS III Off and UPDRS IV scores, highlighting the clinical relevance of precise targeting.ConclusionElectrophysiological sensing, when integrated with imaging data, enhances the iden-tification of optimal STN contacts and supports efficient DBS programming.SignificanceCombining LFP recordings with neuroimaging improves programming accuracy and long-term clinical outcomes, supporting their combined use during STN-DB programming.Keywords: Deep brain stimulation, Subthalamic nucleus, Local field potentials, Neuroimaging, UPDRS Introduction Deep brain stimulation (DBS) of the subthalamic nucleus (STN) suppresses pathological beta-band oscillations (13–30 Hz) and improves motor performance in Parkinson's disease (PD) (Krack et al., 2003; Deuschl et al., 2006; Kühn et al., 2008; Kehnemouyi et al., 2021). The clinical ben-efit of STN-DBS depends heavily on accurate electrode placement and on the initial selection of stimulation parameters (Okun et al., 2005), a process that has traditionally relied on iterative trial-and-error programming guided by clinical response alone. Two largely independent approaches have emerged to make this process more systematic. The first uses neuroimaging-based reconstruction: software such as Lead-DBS combines preoperative MRI and postoperative CT to localize electrode trajectories within standardized brain atlases, allowing direct visualization of contact position relative to the sensorimotor STN (Horn et al., 2019; Zaidel et al., 2009). The second uses electrophysiology: newer implantable systems capable of recording local field potentials (LFP), such as the Medtronic Percept™ PC, allow beta-band activity to be measured directly from the implanted electrode and used as a biomarker for contact selection (Strelow et al., 2022; Feldmann et al., 2022; Krauss et al., 2021; Jimenez-Shahed, 2021). Both approaches have independently improved DBS programming (Nguyen et al., 2020; Segura-Amil et al., 2023; Shah et al., 2023; Fernández-García et al., 2022; Boëx et al., 2023; Busch et al., 2023), but they are rarely validated against each other in the same clinical cohort. It remains unclear to what extent the anatomical “motor sweet spot” identified by imaging coincides with the electrophysiological sweet spot identified by LFP beta peaks, and whether discrepancies between the two methods carry clinical consequences for patient outcomes. This distinction is clinically relevant because imaging requires postoperative acquisition and offline reconstruction, whereas LFP sensing can inform programming during routine follow-up. If the two methods converge on the same contacts, electrophysiology could reduce reliance on imaging-based workflows; if they diverge, this has direct implications for which method should be prioritized in routine practice. This question has not, to our knowledge, been addressed using chronic sensing recordings compared against Lead-DBS reconstructions in parallel within the same patients. Here, we compare neuroimaging-based electrode classification against LFP beta activity recorded with chronic sensing in 59 PD patients who underwent bilateral STN-DBS, and assess how concordance between the two methods relates to motor outcomes at one year, measured with the Unified Parkinson's Disease Rating Scale (UPDRS). Materials and methods Participants All patients were diagnosed with PD according to the MDS Parkinson's disease criteria. This study was conducted at the Cruces University Hospital and Biobizkaia Research Institute. Data from 59 patients with PD who underwent bilateral STN-DBS from May 2021 to May 2024 were in-cluded. All patients were implanted with Percept™ PC implantable pulse generators from Medtronic (Minneapolis, MN). Each patient's motor function was assessed using the Unified Parkinson's Dis-ease Rating Scale (UPDRS) as part of a CAPSIT surgery protocol, before and 6 to 12 months after the surgery. The UPDRS III scale was administered under two standardized conditions: “medica-tion OFF,” after a 12-h withdrawal of dopaminergic medication, and “medication ON,” 1 h after patients received their usual L-Dopa dose. In both postoperative assessments, DBS stimula-tion was active and set to the optimized parameters established during the programming session, i.e. stimulation ON/medication OFF and stimulation ON/medication ON. Additionally, the UP-DRS IV scale was used to evaluate the motor complications before and after surgery. The study was approved by the local ethics committee and all patients signed a written informed consent to participate in the study. The cohort consisted of 23 females and 36 males, with a mean age of 61.67 years (SD = 8.23, range = 33.0 years). The age distribution had a 25th percentile of 56.25 years and a 75th percentile of 68.0 years. The average duration of Parkinson's disease among the patients was 10.43 years (SD = 3.16), with a 25th percentile of 8.0 years and a 75th percentile of 12.75 years. The sex distribution showed a male proportion of 61.01%. Of the 59 patients, 42 (71.2%) presented with an akinetic-rigid phenotype (R-A) and 17 (28.8%) with a tremor-dominant phenotype (T). Sensing testing All patients were implanted with Medtronic SenSight™ directional leads (ref. B33005, Medtronic Inc.) connected to a Percept™ PC implantable pulse generator (IPG), placed abdominally in all cases. Electrode impedance was measured and confirmed to be within normal ranges for all contacts prior to LFP recordings. Although SenSight™ directional leads were implanted in all patients, directional LFP analysis was not performed. Given the sample size, separating recordings by angular segment would have resulted in insufficient statistical power for meaningful comparisons. Directional contacts were therefore treated as omnidirectional for the purposes of this analysis, with only level-based bipolar pairings considered. All patients underwent a BrainSense™ Survey (Medtronic Inc.), a component of the Percept™ PC neurostimulation system (Jimenez-Shahed, 2021) designed to capture local field potentials (LFPs) through the implanted 8-contact DBS lead. This technology enables real-time recording of brain signals in clinical and non-clinical settings while delivering therapeutic stimulation. Fifteen days after surgery, LFP sensing was performed in the clinic in an OFF-medication and OFF-stimulation condition. BrainSense™ Survey was used to collect LFPs from each target through macroelectrode bipolar pairings, which was the only clinically available recording configuration at the time of assessment. Thus, only bipolar recording data were considered for analysis, using the peak of beta-band activity across the entire beta range. Optimal electrode contact selection for stimulation was based on three factors: (1) the best detection of beta-band activity peak (both frequency and amplitude) during the BrainSense™ Survey, (2) the electrode location within the STN according to neuroimaging, and (3) the clinical response during monopolar testing. In all patients, at least one contact fulfilled these criteria to a sufficient degree to guide programming. In cases where beta activity was absent or ambiguous, neuroimaging-based localization and the clinical response during monopolar testing served as the primary criterion for contact selection. Thirty days after the procedure, stimulation was initiated using the contact selected according to this hierarchy. Electrode trajectory reconstruction and location classification All patients adhered to a standardized perioperative protocol, which included a preoperative cranial MRI and a postoperative neuronavigation CT scan conducted 24 h following the inter-vention. DBS electrodes were localized using the Lead-DBS advanced processing pipeline (Horn et al. (2019); RRID:SCR002915; lead-dbs.org). Postoperative CT scans were linearly coregistered to preoperative MRI using ANTs (Avants et al., 2008); coregistrations were visually inspected and manually refined where needed. Preoperative volumes were then used to estimate a nonlinear nor-malization to ICBM 2009b NLIN asymmetric MNI space applying the ANTs SyN Diffeomorphic Mapping using the preset effective: low variance default + subcortical refinement (Avants et al., 2008; Fonov et al., 2009). Electrode contacts were automatically pre-reconstructed using PaCER and electrode trajectories, including axial orientation of the lead, were manually refined and visu-ally validated. The motor territory of the STN was defined according to the DISTAL atlas (Ewert et al., 2018), which parcellates the STN into sensorimotor, associative, and limbic subdivisions. Electrodes were classified as optimal when the active contact was located within the sensorimotor STN as defined by this atlas; suboptimal when the contact was adjacent to but not fully within the sensorimotor territory; and out-STN when no contact was within the nucleus. The DBS motor sweet spot used as an anatomical reference for MNI-to-patient space transformation was derived from Horn et al. (2017) and served to anchor the center of the motor STN target in individual patient space. Statistics Descriptive statistics were calculated for demographic and clinical characteristics of the 59 sub-jects. Electrodes were analyzed individually, with a total of 118 electrodes. The Shapiro-Wilk test was used to determine the normality of beta activity data for frequency and magnitude. Normality confirmed, an ANOVA test was conducted to compare differences in beta activity among three groups, accompanied by multiple comparisons of means using Tukey's HSD test for the significant ANOVA test. To assess the correlation between electrode contact chosen based on beta activity and those chosen based on neuroimaging and clinical criteria, chi-square tests were performed for each hemisphere. For the UPDRS motor study data, which included 33 patients out of the initial 59 after deleating those with no UPDRS data, first, data were divided into two groups; Group A; for electrode pairs (left-right hemisphere) where both electrodes were in an optimal location or at least one of them was optimal while the other hemisphere's electrode was suboptimal, and Group B for the rest of the combinations (out-STN – suboptimal, suboptimal - suboptimal, optimal - out). This subdivision was chosen to ensure two clinically meaningful and comparable groups despite the limited sample size, grouping patients with optimal–suboptimal electrode pairs together and separating those with poorer placement combinations. For a visual representation of what optimal, suboptimal, out-STN means please refer to Fig. 1. Paired t-tests were performed to evaluate improvements within each group at 6 and 12 months postoperatively, following confirmation of normality using the Kolmogorov-Smirnov test. All the analysis code and procedures are publicly available at github.com/compneurobilbao/stn-beta-analysis. Lead-DBS reconstructions illustrating electrode placements across different classifica-tions within the Subthalamic Nucleus (STN). Panels (a) and (b) show optimal electrode placements, demonstrating precise targeting within the STN in different views. Panels (c) and (d) show sub-optimal electrode placements, with electrodes located near the STN but not optimally positioned. Panels (e) and (f) represent out-STN placements, where the electrodes are located outside the STN. Results Correlations between sensing and neuroimaging 118 dB electrodes were analyzed and categorized as previously mentioned (Table 1). While most patients showed a peak in beta activity during the sensing test, some (particularly those clas-sified as suboptimal or located outside the STN) had no detectable beta activity. These recordings were compared to determine whether beta activity levels were higher in magnitude or had a different frequency in optimal placements. Statistical analysis revealed no significant differences in magni-tude (F = 0.49, p = 0.609). However, a significant difference was found in the frequency of beta activity across the three groups (F = 6.06, p = 0.003), indicating that, in this cohort, frequency may be more sensitive to lead placement. This is illustrated in Fig. 2. Table 1Distribution of electrode placement.ClassificationValue CountProportion (%)Optimal8168.64Suboptimal2823.72Out-STN97.62 Distribution of electrode placement. Box plots illustrating the distribution of frequency (Section A) and magnitude (Section B) of neural signals based on neuroimaging-guided electrode location categories: optimal, suboptimal, and out. Data from 118 electrodes. To further explore these group differences, a Tukey HSD post hoc test was performed. The results showed a significant difference in beta frequency between the optimal and suboptimal groups (mean difference = −4.29 Hz, p = 0.0073), while no significant differences were observed between the out–optimal or out–suboptimal comparisons (p = 0.0938 and 0.9849, respectively). However, a significant association was found between the contacts chosen by sensing and those identified by neuroimaging (Lead DBS) (Fig. 3). For the right hemisphere, the chi-square test showed a value of 43.66 with a p-value of 7.54 × 10−9, indicating a strong correlation between the two methods. Similarly, for the left hemisphere, the chi-square test result was 24.89 with a p-value of 5.27 × 10−5, which reinforces the correlation. Frequency of contact selection alignments between the two guidance methods, highlighting differences in electrode targeting accuracy. (A) Left hemisphere, i.e.: selection of contact 1 using sensing matches with neuroimaging in 21 electrodes and in 1 case sensing identified contact 2 and neuroimaging contact 1. (B) Right hemisphere: i.e.: selection of contact 10 using sensing matches with neuroimaging in 25 electrodes and in 1 case sensing identified contact 9 and neuroimaging contact 10. Clinical correlations For this purpose, data were divided into two groups, Group A (optimal/optimal or opti-mal/suboptimal) and Group B (out/suboptimal, suboptimal/suboptimal or optimal/out). High differences in all UPDRS scores were noted between both groups before and after the surgery. These differences are visually represented in Fig. 4 where patients from Group A significantly improved in motor scales after surgery, while patients from Group B did not. Violin plots showing the distribution of Unified Parkinson's Disease Rating Scale (UPDRS) scores before (Pre-DBS, red) and after (Post-DBS, blue) surgery. All postoperative assessments were conducted with DBS stimulation active at optimized parameters. Group A (upper panels, n = 24) includes patients where both hemispheres had optimally placed electrodes, or one hemisphere was optimal and the other suboptimal. Group B (lower panels, n = 9) in-cludes patients with poorer placement combinations (suboptimal/suboptimal, optimal/out-STN, or suboptimal/out-STN). Each row displays three motor outcome measures: UPDRS III assessed in medication OFF state (stimulation ON/medication OFF), UPDRS III assessed in medication ON state (stimulation ON/medication ON), and UPDRS IV (motor complications). Group A shows statistically significant improvement across all three measures following DBS surgery, while Group B does not show significant changes in any of the scores. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Discussion Electrode location and beta activity analysis DBS therapy is a highly effective treatment for patients with PD and proper electrode location is fundamental for a good clinical outcome (Nguyen et al., 2020). Customization of the stimulation parameters to the needs of each patient is a challenge that requires time and expertise. The introduction of newer implantable generators and directional electrodes has substantially increased the number of possible contact and waveform combinations, making manual selection in clinic increasingly difficult for the treating neurologist. Biomarkers that can guide contact selection and assist DBS programming are therefore needed. Previous work has consistently shown that beta-band power is tightly linked to the spatial re-lationship between the recording contact and the dorsolateral sensorimotor STN, with higher beta amplitude and stronger beta suppression during stimulation typically observed near the electrophys-iological and anatomical “motor sweet spot” (Dembek et al., 2019; Busch et al., 2023; Strelow et al., 2022). These studies converge on the view that beta power has been the most robust physiological marker for spatially specific contact selection and optimal programming within the STN. Current programming strategies largely reflect this framework, prioritizing contacts with the highest beta power within or near the motor territory of the nucleus. In our cohort, 68% of electrodes were located in an optimal position within the STN. Beta magnitude did not differ significantly across optimal, suboptimal, and out-of-STN placements, but beta frequency did, with suboptimal electrodes showing lower frequency than optimal ones. This suggests that frequency may capture aspects of lead positioning not reflected in power alone, particularly under chronic recording conditions that differ from intraoperative paradigms. Methodological features of this study may explain this divergence from prior reports emphasizing beta power, and at the same time support the central role of imaging. The strong imbalance between groups (only 9 out-of-STN vs. 81 optimal electrodes), the presence of patients without a clear beta peak, and the exclusive reliance on bipolar recordings can all attenuate detectable power differences and reduce the generalizability of power-based metrics. Under these conditions, LFP features, whether power or frequency, cannot provide a complete description of lead trajectory, contact-to-nucleus geometry, or proximity to adjacent structures. Consequently, the partial mismatch with previous power-centric findings reinforces the idea that electrophysiology should be treated as one component of a multimodal framework. In this work, imaging controls for precise lead localization, and LFP-derived markers are used to refine contact selection and programming rather than replace anatomical guidance. The discrepancy between frequency and magnitude as spatial markers of electrode placement asks for a explicit interpretation. Methodologically, beta magnitude in bipolar recordings is suscep-tible to a geometric confound. When one contact lies within the STN and the adjacent contact falls outside it, the bipolar subtraction captures a steep voltage gradient across the nuclear boundary rather than purely local oscillatory activity, potentially inflating magnitude estimates paradoxi-cally in suboptimal or out-STN placements. Several out-of-STN electrodes show comparatively high magnitudes in Fig. 2, which is consistent with this artifact. This blurs the spatial gradient that magnitude-based contact selection depends on, and may explain why no significant group differences in magnitude were found despite clear anatomical differences in lead placement. This findings potentially explain why magnitude alone might be insufficient for contact selection, and further support the role of neuroimaging in providing the anatomical context needed to interpret LFP recordings. Beta frequency, by contrast, is a spectral property that reflects the dominant rhythmic character of local neural tissue independently of voltage gradients between contacts, being less susceptible to this geometric artifact. Physiologically, the sensorimotor territory of the STN is known to exhibit a distinct beta frequency profile relative to more ventromedial associative and limbic subdivisions, with higher-frequency beta activity (high beta, ∼20–30 Hz) predominating in the dorsolateral motor region and lower-frequency beta (∼13–20 Hz) in more medial territories (Zaidel et al., 2009). The observed mean frequency difference of −4.29 Hz between optimal and suboptimal placements is therefore consistent with electrodes in suboptimal positions sampling neural populations outside the core sensorimotor STN, where lower beta frequencies were detected. These methodological and physiological considerations suggest that beta frequency may serve as a complementary spatial biomarker that captures information about intra-nuclear electrode positioning that magnitude alone cannot reliably provide in bipolar recording configurations. Exploratory analysis of LFP characteristics by motor phenotype revealed no significant differ-ence in beta frequency between tremor-dominant and akinetic-rigid patients (Mann-Whitney U, p = 0.814), suggesting that beta frequency as a spatial marker of electrode placement is independent of clinical phenotype. Beta magnitude, however, was significantly higher in akinetic-rigid patients (mean 2.24 μV vs 1.62 μV, p = 0.012), consistent with prior reports of greater subthalamic beta synchronization in akinetic-rigid compared to tremor-dominant Parkinson's disease. These findings should be interpreted with caution given the sample size imbalance between groups (n = 42 R-A vs n = 17 T), and replication in larger, balanced cohorts is warranted. Significant association between LFP and neuroimaging based electrode contacts The significant association found between LFP-identified contacts and those identified by neu-roimaging in both hemispheres (p < 7.54 × 10−9 for the right and 5.27 × 10−5 for the left) demon-strates a strong concordance between electrophysiological markers and anatomical reconstructions. This aligns with previous work showing that combining LFP-based biomarkers with imaging-derived localization improves the identification of clinically effective DBS contacts (Di Biase et al., 2023; Horn et al., 2019). Importantly, although (Di Biase et al., 2023) examined intra-operative LFP beta-power rather than chronic recordings, their study provides conceptual support for the com-plementary value of electrophysiology and neuroimaging in DBS lead assessment. In our case, the LFP data were obtained from chronic sensing recordings, which have been shown to provide stable detection of beta-band features across postoperative sessions. The strong alignment seen in the contingency tables (particularly the high frequency of agreement in specific contacts) illustrates the consistency and accuracy achievable when these complementary modalities are combined. Clinically, while chronic LFP recordings effectively reveal which of the eight contacts is electrophysiologically nearest to the STN, they alone lack the spatial resolution to delineate the precise trajectory of the lead. This limitation emphasizes the importance of integrating electrophysiological data with post-operative neuroimaging to fully characterize electrode placement and optimize DBS programming. Newer sensing approaches may reduce some of the spatial limitations of bipolar LFP recordings. The BrainSense™ Electrode Identifier (EI) (Stanslaski et al., 2024), which received CE Mark and FDA approval in early 2025, leverages LFP and alpha-beta activity to identify the optimal contact with substantially greater precision and efficiency than standard monopolar review, providing a contact-level resolution that bipolar configurations cannot achieve. Supporting this, a monopo-lar referencing strategy using the Percept SP Advanced Sense research system has been shown to yield significantly greater accuracy in beta peak classification compared to bipolar recordings, with monopolar configurations resulting in lower mean squared error in contact identification (Baker et al., 2024). As such, future work employing monopolar or pseudo-monopolar sensing configura-tions could overcome the limitations of bipolar recording identified in the present study, potentially enhancing the independent contribution of electrophysiological sensing to contact selection and reducing reliance on neuroimaging as the primary localization tool. DBS electrode placement significantly reduces UPDRS scores The significant improvements in UPDRS III and IV scores post-STN-DBS surgery confirm the efficacy of DBS in alleviating motor symptoms and reducing therapy-related complications in PD patients. Group A, characterized by optimal electrode placements, exhibited statistically significant reductions in UPDRS III Off and UPDRS IV scores. The use of the Mann-Whitney U test for non-normally distributed data reinforces the robustness of these findings, confirming that the differences observed between the groups are not due to random variability. These results support the clinical importance of accurate electrode placement. Importantly, all postoperative assessments were conducted with stimulation active at optimized parameters, evaluated under both medication OFF and medication ON conditions. The consistent improvement pattern across both stimulation ON/medication OFF and stimulation ON/medication ON conditions in Group A suggests that optimal electrode placement contributes meaningfully to motor benefit independently of medication state, reinforcing the clinical relevance of precise anatomical targeting. That Group B failed to show significant improvements under either condition further underscores that stimulation alone is insufficient when electrode placement is suboptimal, and that the anatomical accuracy of lead placement appears to be an important contributor to long-term therapeutic outcome. Regarding phenotype outcomes, no significant differences in UPDRS improvement were observed between tremor-dominant and akinetic-rigid patients across any of the three measures assessed (UPDRS III Off: p = 0.935; UPDRS III On: p = 0.194; UPDRS IV: p = 0.729), suggesting that the clinical benefit of STN-DBS is compa-rable between motor phenotypes when electrode placement is adequate. These findings should be interpreted cautiously given the small size of the tremor-dominant subgroup in the UPDRS cohort (n = 9). Several recent studies have explored automated and machine learning-based approaches for STN localization using electrophysiological signals (Hosny, Zhu, Gao and Fu, 2021a, Hosny, Zhu, Su, Gao and Fu, 2021b, Hosny, Zhu, Gao and Fu, 2022, Hosny, Zhu, Gao and Elshenhab, 2024), offering a comple-mentary perspective to the multimodal framework presented here. These approaches demonstrate that electrophysiological signals alone can support automated, data-driven STN localization with promising accuracy. Our study addresses a different but related question. Rather than proposing a classification algorithm, we evaluate whether two independently acquired localization methods, neuroimaging and chronic LFP sensing, agree on contact selection in a real clinical cohort. This validation step is relevant to AI-driven and closed-loop systems as well. Most adaptive algorithms still depend on a reliable definition of the target contact and the physiological signal to track over time, which is precisely what anatomical and electrophysiological concordance establishes. The multimodal framework described here is therefore not an alternative to data-driven programming strategies, but a validation layer that such strategies can build upon. Limitations The detection of LFP using brain sensing was conducted through bipolar recording, which introduces a fundamental constraint on contact selection: the most dorsal and most ventral contacts of the lead are structurally excluded from ever being identified as optimal, since each contact serves as the reference for its neighboring pair rather than as a signal source. In practice, this restricts sensing-based selection to the middle contacts of the lead, as reflected in Fig. 3, where selected contacts cluster around positions 1–2 and 9–10. This substantially limits the discriminative value of LFP sensing for contact selection, since the effective choice is reduced to two mid-shaft contacts regardless of the true anatomy. It also warrants caution when interpreting the concordance between sensing and neuroimaging, as both methods may converge on the same contacts partly by elimination rather than by independent agreement. Additionally, the reported results are based solely on questionnaire data collected at the “level” scale, rather than the more granular “segment” scale. A further limitation concerns the timing of LFP recordings, which were obtained 15 days post-operatively as part of the standard clinical protocol at our center. Recent literature suggests that the optimal window for stable subthalamic LFP recordings is approximately 3–4 weeks to 3 months after implantation, once transient postoperative effects including the stun/microlesion effect and peri‑lead edema have subsided (Feldmann et al., 2025; de Neeling et al., 2026). Early recordings may therefore be subject to dynamic physiological fluctuations that could affect the stability and interpretability of beta-band features. While this timing reflects our institutional protocol and these recommendations were not available at the time of data collection, future studies should consider later recording windows to ensure LFP features more reliably reflect chronic network dynamics. Conclusions This study demonstrates a strong spatial concordance between LFP recordings and neuroimaging-based electrode localization for identifying active contacts in STN-DBS. How-ever, this high concordance between the two methods does not itself establish that both are independently necessary. A significant difference in beta frequency across optimal, suboptimal, and out-of-STN placements has been identified in this study, this suggests that frequency may capture aspects of lead positioning not reflected in power alone. However, and given the aforementioned limitations, neuroimaging retains clear superiority as a spatial localization tool and may alone be sufficient to guide programming decisions in most cases. The added value of LFP sensing in this matter therefore remains to be demonstrated through analyses of independent variance in clinical outcomes, which the current sample size does not allow to address definitively. Sensing offers independent value as DBS technology moves toward closed-loop neuromodulation, where real-time beta suppression serves as a dynamic biomarker for adaptive stimulation rather than a static marker of contact location. In this context, the role of LFP sensing is not to replace anatomical guidance but to complement it by capturing moment-to-moment fluctuations in pathological activity that imaging cannot provide. Our findings support optimal electrode placement as the primary determinant of clinical outcome, with significant improvements in UPDRS III and IV scores observed only in patients with at least one optimally located electrode. Future studies should test whether LFP-derived and imaging-derived measures independently predict DBS outcomes. Author contributions M.R.: conceptualization, formal analysis, software, writing - original draft, review and editing. E.R.G.: conceptualization, data collection, surgical procedure. B.T.: patient assessment, data collection. T.F.V.: patient assessment, data collection. G.B.: data collection, surgical procedure. A.D.: patient assessment, data collection. I.L.: surgical procedure, patient assessment. J.C.G.E.: patient assessment, writing - review. A.E.: methodology, funding acquisition, conceptualization, supervision, project administration, writing - review and editing. M.R.L.: methodology, data collection, funding acquisition, conceptualization, supervision, project administration, writing - review and editing. Declaration of competing interest The authors declare no potential conflicts of interest. Acknowledgement MR is funded by a predoctoral contract from the Department of Education of the Basque Country (Grant No. PRE_2026_2_0265). AE is funded by the Spanish Ministry of Science and Innovation (Grant No. RYC2021-032390-I) and by Ikerbasque: The Basque Foundation for Science, Bilbao, Spain. MRL is funded by the Spanish Instituto de Salud Carlos III (Grant No. PI23/01270). Contributor Information Malen Razkin, Email: razkinmalen@gmail.com. Edurne Ruiz de Gopegui, Email: edurne.gopegui@gmail.com. Beatriz Tijero, Email: beatriz_tijero@hotmail.com. Tamara Fernández del Valle, Email: Tamarafvalle@gmail.com. Gaizka Bilbao, Email: gaizka.bilbaobarandica@osakidetza.eus. Ainara Dolado, Email: AINARA.DOLADOLLORENTE@osakidetza.eus. Imanol Lambarri, Email: IMANOL.LAMBARRISANMARTIN@osakidetza.eus. Juan Carlos Gómez-Estéban, Email: juancarlos.gomezesteban@gmail.com. Asier Erramuzpe, Email: asier.erramuzpe@gmail.com. Marta Ruiz-Lopez, Email: martaruiznrl@gmail.com. References