Acute and Chronic Local Field Potential Recordings in Dystonia—A Systematic Review Authors: Jack Horan, Eoghan Donlon, Aoibheann Gill, Rosalyn Moran, Conor Fearon, Richard A Walsh, Catherine Moran Published: 2026 ABSTRACT AbstractDystonia is a hyperkinetic movement disorder increasingly conceptualized as a disorder of distributed network dysfunction involving the basal ganglia, cortex and cerebellum. Local field potentials (LFPs) recorded from deep brain stimulation (DBS) electrodes provide a unique opportunity to characterize the electrophysiological signatures underlying this network. We performed a systematic review of acute and chronic LFP recordings in dystonia to define common oscillatory patterns and propose a pathophysiological framework. Ninety three studies including 862 patients were analyzed, comprising predominantly globus pallidus internus (GPi) recordings, alongside subthalamic nucleus (STN) and other network nodes. Across dystonia subtypes, increased low‐frequency oscillatory activity (approximately 3–12 Hz) was the most consistent finding, particularly within the GPi, where it often correlated with clinical severity and demonstrated coherence with cortical and muscle activity. Raised activity across other frequencies were inconsistently observed and appeared to reflect specific disease states or phenotypes rather than a unifying biomarker. Oscillatory abnormalities extended beyond the pallidum to include the GPe, STN and cortex, supporting a model of widespread network dysfunction. Both acute and chronic DBS studies demonstrated that stimulation can modulate low‐frequency activity and network coherence, although these changes do not consistently parallel clinical improvement. Emerging chronic sensing technologies highlight temporal variability and longer‐term network dynamics not captured in acute recordings. Overall, dystonia is characterized by abnormal low‐frequency synchronization across distributed motor networks rather than a single pathological oscillation. Chronic‐sensing provides a unique opportunity to study long‐term network dynamics and may support biomarker development for personalized stimulation strategies in dystonia.Keywords: local field potential (LFP), dystonia, chronic LFP, sensing capable, LFP programming Dystonia is a hyperkinetic movement disorder first described by Oppenheim in 1911, with contemporary definitions as having sustained or intermittent muscle contractions causing abnormal movements or postures or both. 1 , 2 , 3 Deep brain stimulation (DBS) targeting the globus pallidus interna (GPi) is an effective treatment for dystonia. Results vary by dystonia subtype, with DBS more effective in idiopathic than acquired dystonia (excluding tardive dystonia). 4 , 5 , 6 Dystonia is a network disorder involving abnormalities in spinal and brainstem reflexes, cortex, basal ganglia and cerebellar regions. 7 , 8 , 9 , 10 Abnormal oscillatory activity has been recognized as a pathophysiological process in several movement disorders. Pathological oscillatory activity was first recognized in Parkinson's Disease (PD), where abnormally increased beta band activity is a reliable biomarker of bradykinesia, that is suppressed by dopaminergic medication or high frequency stimulation (HFS) in tandem with clinical improvement. 11 , 12 , 13 Such oscillations are assessed through local field potentials (LFPs), which reflect the synchronous activity of neuronal populations. 14 , 15 DBS electrodes allow direct recording of LFPs from deep brain structures, providing insight into network dynamics underlying PD and dystonia. 11 , 16 Oscillations in LFPs at different frequencies may be coherent across distinct brain regions such as the subthalamic nucleus (STN) and cortex. Information transfer across distinct brain areas may be a function of paired oscillatory activity. 17 However, how these network abnormalities are reflected in LFPs in dystonia is incompletely understood. Low‐frequency abnormalities have also been implicated in dystonia. Acute postoperative GPi LFP recordings in idiopathic dystonia show excessive low‐frequency (4–10 Hz) and reduced beta activity at rest. 18 Increased power in the 8–20 Hz range decreases during voluntary or passive movement, while voluntary movement increases 30–90 Hz activity, possibly reflecting pro‐kinetic efferent signaling. Hypertonic dystonic states exhibit increased theta, alpha, and low‐beta power (3–20 Hz). 19 In cervical dystonia (CD), two peaks—theta (4–12 Hz) and beta (13–35 Hz)—have been reported, with theta band power correlated with preoperative symptom severity on the Toronto Western Spasmodic Torticollis Rating Scale (TWSTRS). 20 The DBS contact with the highest theta activity showed the greatest correlation with symptom severity. 21 Recordings in mixed idiopathic and acquired dystonia cohorts similarly show abnormal low‐frequency GPi activity that is suppressed by HFS in patients with primary phasic, but not tonic, dystonia. This suppression parallels immediate clinical improvement in several phasic dystonia cases. 22 Previously LFP recordings were only possible intra‐operatively or in the immediate post‐operative period through externalization of DBS leads prior to implantable pulse generator (IPG) insertion. Chronic LFP recordings have been made possible through the development of the sensing capable IPGs. LFP recordings can be performed years following DBS implantation, off or on stimulation and can be set to record significant clinical events or symptoms at home. 23 These systems provide a unique opportunity to study long‐term network dynamics underlying dystonia and other DBS‐treated disorders. We systematically review all published studies examining acute or chronic LFP characteristics in dystonia and propose a conceptual framework for its underlying pathophysiology. Methods A systematic review of the literature on acute and chronic LFP recordings in dystonia was conducted from database inception to end October 2025, following Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) guidelines. The review was registered on PROSPERO (ID: CRD420251229291). Search Strategy PubMed and Embase were searched for English‐language articles using the following terms: (dystonia OR dystonic OR dystonic disorders OR torticollis OR writer's cramp OR DYT OR blepharospasm OR meige) AND (local field potential OR LFP OR low frequency oscillation OR theta OR chronic LFP OR acute LFP OR Percept). Reference lists of studies included were examined for additional eligible articles. Two reviewers independently screened titles and abstracts for suitability. Inclusion/exclusion criteria Studies reporting the following were included: LFPs in dystonia in English language only, LFPs in both dystonia and other disorders, original articles, letters to editors, case reports short communications and pre‐prints relating to adult and pediatric patients. There was no minimum follow‐up time. Acute recordings were defined as intra‐operative or post‐operative recordings while leads were externalized prior to IPG insertion. Chronic recordings were defined as recordings from a sensing enabled IPG beyond the initial post‐operative period. Low‐frequency activity definitions may vary by study but 3–12 Hz encompasses the frequencies covered by the studies included. Exclusion criteria included: LFPs in non‐dystonia studies, reviews without original data, non‐human studies, conference abstracts, studies reporting single unit recordings with no LFP data or unavailable full texts. Studies failing to meet inclusion criteria based on title or abstract were excluded during screening. Quality assessment Two authors (JH and ED) assessed the eligibility of articles, with studies only included if both authors agreed. Disagreements were resolved by discussion with the senior author (CM). Risk of bias was evaluated using the ROBINS‐E (Risk Of Bias in Non‐Randomized Studies—of Exposures) tool (Table S1). 24 Results Ninety three articles with LFPs recorded from 862 patients with dystonia were included (Table 1). 830 patients had bilateral GPi or STN leads. 13 patients from two studies had more than two leads inserted (one with rescue STN, 12 patients up to 12 temporary leads), 19 patients from three studies had temporary multi‐site recordings. A PRISMA flowchart is shown (Figure 1). Many studies reported multiple subtypes of dystonia. Dystonia subtypes included are shown in Table 2 (cervical, non‐specified, generalized and Meige syndrome were the most common subtypes included). Non specified were a mixed group of idiopathic adult onset dystonia of various patterns. Acute recordings were recorded in 78 studies in 800 patients. Chronic recordings from 15 studies in 62 patients were included. TABLE 1Included studies investigating local field potential characteristics in dystonia and study main findingsStudyYearPatient numberDystonia typeAdult/pediatricRecording siteRecording periodRecording TimeTestingFrequency of interestMain findingLiu 70 20021Familial myoclonic dystoniaAdultGPi (bil)AcuteIntra‐op At Rest Movement 2–8 Hz peakIncreased oscillation at 4–8 Hz during myoclonic activity, coherent with EMG muscle activityNandi 87 20021Segmental dystonia (tardive camptocormia)AdultGPi (bil)Acute Post‐op 2 days At Rest Movement 0–20 Hz increased activityLFP and EMG coherence during muscle activity at 0.5–3.5 HzSilberstein 18 2003 4 1 4 1 12 Generalized Segmental Cervical Multisegmental PD AdultGPi (bil)Acute Post‐op 2–7 days At Rest4–10 Hz increased activityRaised 4–10 Hz in dystonia compared to PDChen 31 200613Dystonia with cervical involvementAdultGPi (bil)Acute Post‐op 2–6 days At Rest Movement Neck posturing 4–10 Hz increased activityCoherence between 4–10 Hz LFP recording and dystonic EMGChen 35 2006 2 3 1 Generalized Segmental Cervical AdultGPi (bil)AcuteIntra‐opAt Rest3–12 HzLFP oscillations are highest in GPi in the 3–12 Hz rangeLiu 69 2006 5 1 2 Generalized Segmental Cervical AdultGPi (bil)Acute Post‐op 4–6 days At Rest Movement 4–12 Hz increased activityCoherence between LFP recording and rhythmic dystonic EMG activityFoncke 71 20072Familial myoclonic dystoniaAdultGPi (bil)AcuteN/A At Rest Movement Go‐No‐Go Task 3–15 Hz increased activityCoherence between LFP recording and EMG activity 3–15 Hz at rest with desynchronization following taskMarceglia 36 2007 4 1 Dystonia PD Adult GPi (bil) GPi/STN Acute Post‐op 2 days At Rest4–14 Hz increased activityRaised low frequency activity in dystoniaTang 65 20072CervicalAdultGPi (bil)Acute Post‐op 2–3 days At Rest Sensory Task 6–8 Hz reduced activity with sensory trickDuring effective sensory trick desynchronization of raised 6–8 Hz activityBrücke 88 2008 1 2 4 4 Generalized Generalized myoclonic Segmental Focal AdultGPi (bil)Acute Post‐op 2–6 days At Rest Motor Task 18–25 Hz reduced activity, 60–80 Hz increased activity with movement Increase in gamma activity (60–80 Hz) during movement of contralateral limb and suppression of low frequency activityKühn 37 2008 3 1 2 Generalized Multifocal Tardive AdultGPi (bil)Acute Post‐op 2–6 days At Rest5–12 Hz increased activityRaised low frequency activity at rest. Given dopamine inhibitor, increase in low beta (13–20 Hz) activityLiu 19 2008 9 6 Generalized Cervical AdultGPi (bil)Acute Post‐op 4–6 days At Rest Movement 3–18 HzDecreased 8–20 Hz with voluntary, passive movement and vibrationSharott 68 20087 Mixed group (primary), all with neck involvement AdultGPi (bil)Acute Post‐op 2–6 days At Rest Movement Below 10 HzPallidal LFP below 10 Hz is coherent with and drives dystonic EMG muscle activityHuberdeau 27 20112CervicalAdultGPi (bil)AcuteIntra‐opAt RestLow frequency peaks 6–10 HzRaised low frequency activity at restSingh 89 2011 1 5 2 Segmental Cervical Meige AdultGPi (bil)Acute Post‐op 1–2 days At Rest Movement During walking increased theta, alpha and gamma frequency power with decreased beta powerSingh 90 2011 8 9 Cervical PD Adult GPi (bil) STN (bil) Acute Post‐op 2–4 days At Rest Movement 8–12 HzRapid complex arm movements produced increased alpha frequency activity in dystonia and PDBrücke 91 2012 3 18 1 Segmental Cervical Focal AdultGPi (bil)Acute Post‐op 1–4 days Reaction time task35–105 HzDuring movement in task, larger and quicker movements produced increased gamma synchronizationNeumann 59 20121GeneralizedAdultSTN (bil)Acute Post‐op 2–5 days At Rest7 Hz peakRaised low frequency (4–12 Hz) is present in STN in dystoniaTsang 92 20125CervicalAdultGPi (bil)Acute Post‐op 1–3 days At Rest Movement Task 10–24 Hz Movement related GPi beta desynchronization (10–24 Hz) and gamma synchronization (64–68 Hz) Coherence between GPi was present at rest at 5–18 Hz Weinberger 32 2012 1 1 5 4 Generalized Segmental Cervical PD Adult GPi (bil) GPi (bil) Acute Acute Intra‐op Intra‐op At Rest At Rest 8–20 Hz 11–30 Hz GPi oscillatory activity differed in dystonia versus PDSingh 93 2013 2 2 2 1 Generalized Segmental Cervical Tourette's AdultGPi (bil)Acute Post‐op 2–3 days At Rest Movement Task 15–80 HzLow beta activity decreased while high beta and gamma activity increased on contralateral side during movement taskWhitmer 44 2013 4 1 Secondary dystonia DYT3 Adult GPi (bil) GPi (bil) Acute Acute Intra‐op At Rest During HFS 13–35 HzResting state beta frequency oscillatory activity was present in secondary and DYT3 dystonia. HFS modulated increased alpha and beta power in DYT3Barow 22 2014 1 4 5 1 1 Generalized Segmental Cervical Tardive Acquired (generalized) AdultGPi (bil)Acute Post‐op 1–3 days At Rest During HFS Post HFS 4–12 Hz increased activityLow frequency (4–12 Hz) oscillatory activity is reduced with HFS in phasic dystonia. Decreased coherence with GPi LFP and cortical EEG and muscle EMG post HFSHerrojo Ruiz 94 2014 8 2 Cervical DYT11 AdultGPi (bil)Acute Post‐op 2–3 days Sensorimotor Task14–30 HzDuring sensorimotor task, dynamic changes are present in GPi beta band LFPHerrojo Ruiz 63 2014 6 2 1 Cervical Segmental Orofacial AdultGPi (bil)Acute Post‐op 2–4 days Flanker Task4–8 HzTheta activity increased during error response. Pallidal theta increased before cortical theta suggesting GPi to cortical directionalityLee 53 201411CervicalAdultGPi (bil)AcuteIntra‐opAt Rest4–12 HzInterhemispheric differences in low frequency (4–12 Hz) and beta (13–30 Hz) activity in CD with rotation and laterocollisMoll 28 201413CervicalAdultGPi (bil)AcuteIntra‐op At Rest Movement 2–8 HzTheta band prominence in GPi and GPe in CD, with interhemispheric differences in spectral powerFranzini 95 20141Generalized (acquired)PediatricGPi (bil)AcutePost‐op Day 12At RestPatient had GPi DBS and underwent pallidotomy. LFP recordings before and after lesioning. Reduction in LFP amplitudes after pallidotomyMamun 96 2015 6 6 Dystonia PD Adult GPi (bil) STN (bil) Acute Post‐op 4–6 days Finger‐click TaskLFP analysis could accurately decode movement related behaviorNeumann 72 2015 2 2 5 Generalized Segmental Cervical AdultGPi (bil)Acute Post‐op 1–7 days At Rest5–95 HzCoherence between GPi and cortical areas were identified across different frequencies. Theta – pallidotemporal, alpha – pallidocerebellar and beta—corticopallidalWelter 41 2015 6 6 Generalized Myoclonus dystonia AdultGPi (bil)AcuteIntra‐opAt Rest4–8 HzRaised theta oscillatory activity was present in generalized and myoclonus dystoniaTalakoub 97 2016 1 4 1 1 6 Generalized Cervical Focal Meige PD AdultGPi (bil)Acute Post‐op 1–3 days Movement Based Task 20–30 Hz 65–80 Hz Motor cortex and GPi or STN coherence was coupled during movement in beta and gamma bandsTrenado 29 2016 7 5 Cervical Tardive dystonia AdultGPi (bil)AcuteIntra‐op At Rest Sensory palmar stimulation 1–7 HzIncreased thetadelta LFP for CD and TD. Reduced peak frequency with SPS in CDWang 57 2016 6 3 3 28 Generalized Craniocervical Cervical PD Adult STN (bil) STN (bil) AcuteIntra‐opAt Rest21 ± 7 HzNo differences in power between dystonia and PD across multiple frequencies were seen. Beta band activity not specific to PDBočková 98 2017 1 1 1 2 5 Generalized Multifocal Orofacial Cervical PD Adult GPi (bil) STN (bil) AcutePost‐opVisuomotor TaskDuring visuomotor task with increased complexity, an event related beta desynchronization was seen in GPi and STNGeng 38 2017 3 4 1 12 Generalized Craniocervical Tardive PD Adult STN (bil) STN (bil) Acute Post‐op 1–5 days At Rest0–10 HzDystonia had higher low frequency and lower beta frequency activity than PD. Medication on PD had lower beta power activityGillies 64 2017 3 2 3 Cervical Focal PD AdultGPi (bil)Acute Post‐op 1–7 days Decision TaskGPi responds differently to correct and incorrect feedback in dystonia and PD.Giorni 99 2017 3 4 7 Cervical Mixed (idiopathic and acquired) Tourette's Adult GPi (bil) GPi (bil) GPi (bil) AcuteIntra‐opAt Rest4–12 HzHigher proportion of alpha (4–12 Hz) activity was present in dystonia compared to TSIslam 100 2017 6 6 Dystonia PD AdultGPi (bil)Acute Post‐op 4–6 days At Rest Movement Task A model was developed to decode movement based on LFP activityNeumann 21 201727CervicalAdultGPi (bil)Acute Post‐op 2–7 days At Rest4–12 HzTheta power was associated with pre‐operative symptom severity in CDTsiokos 101 2017 3 19 Dystonia PD AdultGPi (bil)AcuteIntra‐op At Rest Movement Resting beta low gamma phase amplitude coupling was inversely correlated to bradykinesia and rigidityvan Wijk 102 2017 2 2 4 1 Generalized Segmental Cervical Meige AdultGPi (bil)Acute Post‐op 1–7 days At Rest Reaction Time Task 13–21 HzMovement related decrease in coherence in beta band (13–21 Hz) between GPi LFP and cortical MEGWang 39 2018 7 4 3 20 Generalized Craniocervical Cervical PD Adult GPi (bil) GPi (bil) AcuteIntra‐op At Rest Movement Task 4–8 HzAt rest, theta power is increased in dystonia, with beta power increased in PD. Coherence between motor cortex and GPi was seem in theta frequency in dystoniaYokochi 103 2018 14 1 1 2 1 Generalized Segmental Cervical Tardive Myoclonus AdultGPi (bil)AcuteIntra‐opAt Rest 2–4 Hz 8–13 Hz Phasic dystonia was associated with alpha spectral peaks and alpha coherence between GPi and motor cortex. Tonic dystonia was associated with delta oscillatory activityZhu 30 2018 9 7 Cervical Huntington's disease Adult GPi (bil) GPi (bil) AcuteIntra‐opAt Rest6–14 HzLow frequency activity was present in CD. Oscillatory activity in high beta and low gamma (26–43 Hz) was seen in HDCao 47 20191Acquired dystoniaAdultSTN (bil)Acute Post‐op 1–7 days At Rest 4–12 Hz 28–35 Hz STN – cortical coherence was seen in the 28–35 Hz band. Right STN had coherence in the theta‐alpha band and prefrontal cortexLofredi 34 2019 1 1 2 1 5 Segmental Multifocal Cervical Meige PD AdultGPi (bil)Acute Post‐op 1–7 days At Rest4–8 HzIncreased theta power was present in pallidal LFP in dystonia. Beta bursting characteristics were comparable between dystonia and PD on medicationPiña‐Fuentes 33 2019 2 1 1 1 1 1 11 Segmental Cervical Meige Myoclonus Generalized (acquired) Segmental (acquired) PD Adult GPi (bil) GPi (bil) or STN (bil) AcuteIntra‐opAt Rest4–12 HzGPi low frequency and beta oscillations are different between dystonia and PDScheller 76 2019 3 3 2 1 Generalized Segmental Cervical Tardive AdultGPi (bil) Chronic ACTIVA Post‐op 18.9 +/−13.9 months At Rest 3–12 Hz 13–30 Hz Following months of chronic stimulation, stimulation was turned off. Pallidal low frequency activity was associated with symptom severitySedov 104 201915CervicalAdultGPi (bil)AcuteIntra‐opAt RestOscillatory and single unit activity was asymmetric in CDSedov 51 201912CervicalAdultGPi (bil)AcuteIntra‐opAt RestAsymmetrical neck tone with laterocollis was associated with asymmetrical LFP oscillatory activity and single unit spontaneous discharge patternPiña‐Fuentes 73 2020 1 7 1 1 2 Generalized Segmental Cervical Oromandibular Meige AdultGPi (bil)AcuteIntra‐op At Rest During HFS 4–12 HzLow frequency oscillatory activity is raised in dystonia. This was lower in patients treated with stimulation than stimulation naïve patients. Stimulation suppresses low frequency activityZhu 105 2020 2 4 Generalized (TWNK mutation) Generalized (no TWNK mutation) Adult GPi (bil) STN (bil) Acute Post‐op 2–5 days At Rest15–27 HzHigh beta oscillatory activity was increased in TWNK patients (clinically presents as bradykinesia and rigidity) compared to non‐TWNK patientsJohnson 83 20211CervicalAdultSTN (bil) Chronic SUMMIT Post‐op 2–12 weeks At Home3–7 HzChronic recordings at home. 3–7 Hz band coherent between STN and cortex. Theta band modulated with stimulationSemenova 52 20219CervicalAdultGPi (bil)AcuteIntra‐opAt Rest12–30 HzLFP analysis showed pallidal asymmetry in CD, with pallidum contra‐lateral to head turn more inhibitedThenaisie 23 2021 3 2 14 1 Cervical Myoclonus PD Chronic pain Adult GPi (bil) STN (bil) Chronic PERCEPT Post‐op 2 days–8 years At Rest5.7 Hz peakTheta‐alpha peak present in all GPi in dystonia. Beta peak present in 19/22 STN in PDFasano 84 2022 1 1 Generalized Epilepsy Pediatric Adult GPi (bil) ANT (bil) Chronic PERCEPT Post‐op Days 1.95 peakAnalysis of LFP data from a sensing enabled IPG may aid in programming stimulation settings in dystoniaGiorni 74 20221DYT6 dystoniaAdultGPi (bil)AcuteIntra‐opAt RestGamma band LFP power were lower in a revision surgery 4 years after index surgeryNavid 62 2022 9 8 Isolated dystonia PD Adult GPi (bil) STN (bil) Acute Post‐op 0–2 days Flanker Task3–7 HzEvent related synchronization in theta was seen in GPi and STN during a conflict task. GPi is also involved in cognitive processing in conflict in addition to the STNAverna 106 20232GeneralizedAdultGPi (bil)Acute Post‐op 0–2 days At Rest Movement 4–12 HzRaised low frequency was observed at rest. Voluntary movements increased 4–7 Hz activityCajigas 48 20233Dyskinetic cerebral palsyPediatric and adultDN (bil) Chronic PERCEPT Post‐op 1–3 months At Home9.8 HzProminent alpha activity was observed in the DN of these acquired dystonia patients. Decreased variability in alpha was seen with stimulationChoubdar 107 2023 1 1 Cervical PD AdultGPi (bil)Acute Post‐op 0–7 days Decision Making Task100–150 HzHigh gamma activity present in dystonia after feedback in task, observed in both GPi but not seen in PDDel Vecchio 108 2023 1 4 3 Multifocal Cervical Myoclonus AdultGPi (bil) Chronic PERCEPT Post‐op 6 months At Rest3–12 HzPercept IPG can record low frequency oscillatory activity in chronic stimulation. Sensory trick reduced power in 3–12 Hz band (theta‐alpha)Hernandez‐Martin 60 20239Dyskinetic cerebral palsyPediatric GPi (bil) STN (bil) VOA/VOP (bil) Acute Post‐op 1–2 days Movement Task13–35 HzBeta frequency peaks in GPi, VOA/VOP and STN during movement but not rest. Abnormal patterns of inhibition and disinhibition, instead of reduced GPi activity may contribute to dystoniaJoshi 67 202317CervicalAdultGPi (bil)AcuteIntra‐op At Rest Shoulder movement Proprioception 4–12 HzGPi – cerebellar coherence modulated by neck vibration in isolated CD in alpha bandKasiri 61 2023 1 2 DYT1 KMT2B Pediatric GPi (bil) STN (bil) VOA/VOP (bil) VA VIM AcutePost‐opMovement Task1–35 HzGPi and STN activity increased during movement task. GPi output pattern and not average firing rates abnormal in dystoniaLofredi 109 2023 3 2 1 Generalized Segmental Cervical AdultGPi (bil) Chronic ACTIVA Post‐op At Rest Finger‐tapping Exercise 13–30 HzPallidal stimulation was stopped after a period of effective stimulation. An increase in beta activity was inversely related to velocity. As the time from stimulation increased, so did velocity, with a decrease in beta activitySedov 66 2023 10 7 Generalized Cervical AdultGPi (bil)AcuteIntra‐op At Rest Shoulder Movement Proprioception 4–13 HzNeck vibration showed an increase in burst neuron presence and oscillatory activity in GPi. In CD, alpha band activity in particular was increased with vibrationWiest 58 20237Craniocervical dystoniaAdultSTN (bil)Acute Post‐op 4–7 days At Rest HFS Raised theta activity was not seen in STN in dystonia. Broadband suppression was seen with stimulationYin 110 2023 7 13 11 8 Cervical Meige PD Huntington's AdultGPi (bil)Acute Post‐op 0–5 days During Sleep0–13 HzGPi and GPi‐cortical coherence are impacted by sleep–wake cycle. Sleep states can be decoded by LFP analysisZhang B 40 202328MeigeAdultGPi (bil)AcuteIntra‐opAt Rest4–8 HzAbnormally increased oscillatory activity in theta (4–8 Hz) was present in Meige syndromeZhang R 111 2023 13 8 Generalized Cervical AdultGPi (bil)Acute Post‐op 2–5 days At Rest Muscle Contraction 3–12 HzRaised oscillatory power in theta and alpha bands were seen. Power ratio between frequency bands was altered in high or low muscle contraction and correlated with symptom severityEbden 50 202410 Mixed dystonia (idiopathic and acquired) PediatricGPi (bil) Chronic PERCEPT Post‐opAt Rest0–12 HzOscillatory changes followed a circadian rhythm. Oscillatory power in delta/theta and gamma increased during dystonic symptom periodsGeng 112 2024 1 7 5 Generalized Cranial Cervical Adult GPi (bil) STN (bil) Acute Post‐op 3–5 days At RestCombined bilateral GPi and STN DBS. GPi theta power, STN beta and GPi‐STN coupling in theta and beta bands correlated with dystonic symptomsJohnson 113 20241DYT1Pediatric GPi STN AcuteIntra‐op At Rest During HFS 200–500 HzLead revisions, simultaneous recordings from GPi and STN showed ERNA following HFS in GPe and STN, localized to GPe‐STN circuitLofredi 49 2024 4 2 4 Generalized Segmental Cervical AdultGPi (bil)Acute Post‐op 1–7 days At Rest3–12 HzLow frequency oscillatory activity correlated with symptom severity. Coupling between striatum and GPi also correlated with dystonia severitySemenova 114 2024 12 6 3 2 Generalized Segmental Cervical Multifocal AdultGPi (bil)Acute Intra‐op Post‐op Day 1 At Rest During Vibration During HFS Symptom severity was associated with narrowing and asymmetry of multifractal spectrum. This multifractal spectrum became wider with HFSYin 115 2024 48 23 28 11 4 7 Dystonia Dystonia PD Huntington's Tourette's ET Adult STN (bil) GPi (bil) GPi/STN GPi (bil) GPi (bil) Mixed Acute Post‐op 3–7 days During SleepUsing LFP analysis a model was developed to decode sleep stages in movement disorders that has the potential to provide sleep stage specific stimulationAl‐Zubaidi 116 2025 3 8 3 Dystonia PD Tourette's Adult GPi (bil) STN (bil) GPi (bil) AcutePost‐opAuditory Task12–30 HzBeta band power was reduced with onset and duration of words during auditory task in the STN and GPiBalachandar 46 202510SecondaryPediatricGPi (bil) Chronic PERCEPT Post‐op 11–1155 days During status dystonicus13–30 HzPediatric patients implanted with GPi DBS with medically refractive status dystonicus. Increased beta power and burst amplitude during status dystonicusBrinkerhoff 117 2025 1 22 Generalized PD Adult GPi (bil) GPi (bil) or STN(bil) AcuteIntra‐opDuring HFSStimulation delivered intra‐operatively to the STN or GPi showed short‐term neuroplasticityCernera 77 20251CervicalAdultGPi (bil) Chronic SUMMIT Post‐op 0–27 months Daily Life 3–12 Hz 60–65 Hz Chronic recording of LFP in CD. Increased low frequency power (3–12 Hz) and gamma oscillation (60 Hz) off stimulation. Stimulation suppressed low frequency activity and induced gamma oscillations around 65 HzDonlon 45 20252DYT3AdultGPi (bil) Chronic PERCEPT Post‐op 0–12 months At Rest During HFS 13–30 HzElevation of beta band activity was present in one patient with DYT3 dystonia that was progressively suppressed with HFS that correlated with improvement in clinical severityGing‐Jehli 118 2025 3 14 Dystonia PD Adult GPi (bil) STN (bil) AcuteIntra‐opConflict Task4–8 HzBasal ganglia theta activity plays a different role in different basal ganglia nuclei during conflictGuan‐Yu 56 2025 9 17 19 Generalized Cervical Meige AdultSTN (bil)Acute Post‐op 2–4 days At Rest 4–13 Hz 20–30 Hz Low frequency activity is present and localized to sensorimotor area in STN in CD, GD and MS. Low frequency activity correlated with symptoms in CD and GD but not MS. High beta activity was present in STN in all three subtypesHubers 119 20251GNAO1‐related dystoniaPediatricGPi (bil) Chronic PERCEPT Post‐op 0–22 days At Rest7–12 HzLow frequency oscillatory activity was associated with symptom severityKasiri 120 202512Idiopathic and acquiredPediatric and adultMultiple targetsAcutePost‐opAt RestThis study showed the important relationship between structural and functional connectivity in pediatric dystoniaKasiri 75 2025 2 2 1 1 1 Dyskinetic cerebral palsy Mitochondrial disease GA1 KMT2B Unknown Pediatric and young adult GPi (bil) STN (bil) VOA/VOP (bil) VA VIM Acute Post‐op 1–4 days Movement Task During HFS HFS propagates via physiological pathways along axons in mainly orthodromic directionKaymak 42 2025 3 6 1 2 2 2 4 8 3 Cranial Cervical Truncal GNAL KMT2B SCGE (DYT11) THAP1 (DYT6) TOR1A (DYT1) VPS16 AdultGPi (bil)AcuteIntra‐opAt Rest4–12 HzGenetic dystonias had different electrophysiological patterns compared to non‐genetic dystonia. Myoclonus dystonia had lower theta and alpha activity versus idiopathic dystonia and DYT6Kelbert 43 20251DYT1PediatricGPi (bil) Chronic PERCEPT Post‐opIn Clinic11–27 HzBrainsense survey in DYT1 dystonia showed two peaks, a low beta (13.67 Hz) and high beta (25.39 Hz) during initial programmingLedingham 78 20251CervicalAdultGPi (bil) Chronic PERCEPT Post‐op 10–16 weeks In Clinic6.84 HzBrainsense survey at activation showed multiple peaks (theta and beta), with the greatest power in the theta band. Stimulation contacts were chosen based on maximal low frequency LFP powerMünte 121 2025 2 4 Dystonia Tourette's Adult GPi (bil) CM/VOI (bil) AcutePost‐opGo‐No‐Go task12–25 HzDifferent LFP activity was seen for Go and NoGo conditions in the GPi and thalamus, specifically in beta activity. Propose GPi has proactive motor inhibitionPascual 79 20253GeneralizedAdultGPi (bil) Chronic PERCEPT Post‐opIn Clinic5–55 HzOverstimulation produced a simultaneous worsening of clinical symptoms and increase in LFP power spectral density magnitudeAbbreviations: aDBS, adaptive deep brain stimulation; CD, cervical dystonia; CM/VOI, center median/ventralis oralis internus; DN, dentate nucleus; EEG, electroencephalography; EMG, electromyography; GA1, glutaric aciduria 1;GD, Generalized Dystonia; GPe, external globus pallidus; GPi, internal globus pallidus; HD, Huntington's Disease; HFS, high frequency stimulation; IPG, implantable pulse generator; LFP, local field potential; MS, Meige Syndrome; PD, Parkinson's Disease; SPS, sensory palmar stimulation; STN, subthalamic nucleus; TD, tardive dystonia; TS, Tourette's syndrome; VA, ventral anterior nucleus thalamus; VIM, ventral intermediate nucleus thalamus; VOA/VOP, ventral oralis anterior/posterior nucleus thalamus. Included studies investigating local field potential characteristics in dystonia and study main findings At Rest Movement Post‐op 2 days At Rest Movement 4 1 4 1 12 Generalized Segmental Cervical Multisegmental PD Post‐op 2–7 days Post‐op 2–6 days At Rest Movement Neck posturing 2 3 1 Generalized Segmental Cervical 5 1 2 Generalized Segmental Cervical Post‐op 4–6 days At Rest Movement At Rest Movement Go‐No‐Go Task 4 1 Dystonia PD GPi (bil) GPi/STN Post‐op 2 days Post‐op 2–3 days At Rest Sensory Task 1 2 4 4 Generalized Generalized myoclonic Segmental Focal Post‐op 2–6 days At Rest Motor Task 18–25 Hz reduced activity, 60–80 Hz increased activity with movement 3 1 2 Generalized Multifocal Tardive Post‐op 2–6 days 9 6 Generalized Cervical Post‐op 4–6 days At Rest Movement Mixed group (primary), all with neck involvement Post‐op 2–6 days At Rest Movement 1 5 2 Segmental Cervical Meige Post‐op 1–2 days At Rest Movement 8 9 Cervical PD GPi (bil) STN (bil) Post‐op 2–4 days At Rest Movement 3 18 1 Segmental Cervical Focal Post‐op 1–4 days Post‐op 2–5 days Post‐op 1–3 days At Rest Movement Task Movement related GPi beta desynchronization (10–24 Hz) and gamma synchronization (64–68 Hz) Coherence between GPi was present at rest at 5–18 Hz 1 1 5 4 Generalized Segmental Cervical PD GPi (bil) GPi (bil) Acute Acute Intra‐op Intra‐op At Rest At Rest 8–20 Hz 11–30 Hz 2 2 2 1 Generalized Segmental Cervical Tourette's Post‐op 2–3 days At Rest Movement Task 4 1 Secondary dystonia DYT3 GPi (bil) GPi (bil) Acute Acute At Rest During HFS 1 4 5 1 1 Generalized Segmental Cervical Tardive Acquired (generalized) Post‐op 1–3 days At Rest During HFS Post HFS 8 2 Cervical DYT11 Post‐op 2–3 days 6 2 1 Cervical Segmental Orofacial Post‐op 2–4 days At Rest Movement 6 6 Dystonia PD GPi (bil) STN (bil) Post‐op 4–6 days 2 2 5 Generalized Segmental Cervical Post‐op 1–7 days 6 6 Generalized Myoclonus dystonia 1 4 1 1 6 Generalized Cervical Focal Meige PD Post‐op 1–3 days 20–30 Hz 65–80 Hz 7 5 Cervical Tardive dystonia At Rest Sensory palmar stimulation 6 3 3 28 Generalized Craniocervical Cervical PD STN (bil) STN (bil) 1 1 1 2 5 Generalized Multifocal Orofacial Cervical PD GPi (bil) STN (bil) 3 4 1 12 Generalized Craniocervical Tardive PD STN (bil) STN (bil) Post‐op 1–5 days 3 2 3 Cervical Focal PD Post‐op 1–7 days 3 4 7 Cervical Mixed (idiopathic and acquired) Tourette's GPi (bil) GPi (bil) GPi (bil) 6 6 Dystonia PD Post‐op 4–6 days At Rest Movement Task Post‐op 2–7 days 3 19 Dystonia PD At Rest Movement 2 2 4 1 Generalized Segmental Cervical Meige Post‐op 1–7 days At Rest Reaction Time Task 7 4 3 20 Generalized Craniocervical Cervical PD GPi (bil) GPi (bil) At Rest Movement Task 14 1 1 2 1 Generalized Segmental Cervical Tardive Myoclonus 2–4 Hz 8–13 Hz 9 7 Cervical Huntington's disease GPi (bil) GPi (bil) Post‐op 1–7 days 4–12 Hz 28–35 Hz 1 1 2 1 5 Segmental Multifocal Cervical Meige PD Post‐op 1–7 days 2 1 1 1 1 1 11 Segmental Cervical Meige Myoclonus Generalized (acquired) Segmental (acquired) PD GPi (bil) GPi (bil) or STN (bil) 3 3 2 1 Generalized Segmental Cervical Tardive Chronic ACTIVA Post‐op 18.9 +/−13.9 months 3–12 Hz 13–30 Hz 1 7 1 1 2 Generalized Segmental Cervical Oromandibular Meige At Rest During HFS 2 4 Generalized (TWNK mutation) Generalized (no TWNK mutation) GPi (bil) STN (bil) Post‐op 2–5 days Chronic SUMMIT Post‐op 2–12 weeks 3 2 14 1 Cervical Myoclonus PD Chronic pain GPi (bil) STN (bil) Chronic PERCEPT Post‐op 2 days–8 years 1 1 Generalized Epilepsy Pediatric Adult GPi (bil) ANT (bil) Chronic PERCEPT Post‐op Days 9 8 Isolated dystonia PD GPi (bil) STN (bil) Post‐op 0–2 days Post‐op 0–2 days At Rest Movement Chronic PERCEPT Post‐op 1–3 months 1 1 Cervical PD Post‐op 0–7 days 1 4 3 Multifocal Cervical Myoclonus Chronic PERCEPT Post‐op 6 months GPi (bil) STN (bil) VOA/VOP (bil) Post‐op 1–2 days At Rest Shoulder movement Proprioception 1 2 DYT1 KMT2B GPi (bil) STN (bil) VOA/VOP (bil) VA VIM 3 2 1 Generalized Segmental Cervical Chronic ACTIVA At Rest Finger‐tapping Exercise 10 7 Generalized Cervical At Rest Shoulder Movement Proprioception Post‐op 4–7 days At Rest HFS 7 13 11 8 Cervical Meige PD Huntington's Post‐op 0–5 days 13 8 Generalized Cervical Post‐op 2–5 days At Rest Muscle Contraction Mixed dystonia (idiopathic and acquired) Chronic PERCEPT 1 7 5 Generalized Cranial Cervical GPi (bil) STN (bil) Post‐op 3–5 days GPi STN At Rest During HFS 4 2 4 Generalized Segmental Cervical Post‐op 1–7 days 12 6 3 2 Generalized Segmental Cervical Multifocal Intra‐op Post‐op Day 1 At Rest During Vibration During HFS 48 23 28 11 4 7 Dystonia Dystonia PD Huntington's Tourette's ET STN (bil) GPi (bil) GPi/STN GPi (bil) GPi (bil) Mixed Post‐op 3–7 days 3 8 3 Dystonia PD Tourette's GPi (bil) STN (bil) GPi (bil) Chronic PERCEPT Post‐op 11–1155 days 1 22 Generalized PD GPi (bil) GPi (bil) or STN(bil) Chronic SUMMIT Post‐op 0–27 months 3–12 Hz 60–65 Hz Chronic PERCEPT Post‐op 0–12 months At Rest During HFS 3 14 Dystonia PD GPi (bil) STN (bil) 9 17 19 Generalized Cervical Meige Post‐op 2–4 days 4–13 Hz 20–30 Hz Chronic PERCEPT Post‐op 0–22 days 2 2 1 1 1 Dyskinetic cerebral palsy Mitochondrial disease GA1 KMT2B Unknown GPi (bil) STN (bil) VOA/VOP (bil) VA VIM Post‐op 1–4 days Movement Task During HFS 3 6 1 2 2 2 4 8 3 Cranial Cervical Truncal GNAL KMT2B SCGE (DYT11) THAP1 (DYT6) TOR1A (DYT1) VPS16 Chronic PERCEPT Chronic PERCEPT Post‐op 10–16 weeks 2 4 Dystonia Tourette's GPi (bil) CM/VOI (bil) Chronic PERCEPT Abbreviations: aDBS, adaptive deep brain stimulation; CD, cervical dystonia; CM/VOI, center median/ventralis oralis internus; DN, dentate nucleus; EEG, electroencephalography; EMG, electromyography; GA1, glutaric aciduria 1;GD, Generalized Dystonia; GPe, external globus pallidus; GPi, internal globus pallidus; HD, Huntington's Disease; HFS, high frequency stimulation; IPG, implantable pulse generator; LFP, local field potential; MS, Meige Syndrome; PD, Parkinson's Disease; SPS, sensory palmar stimulation; STN, subthalamic nucleus; TD, tardive dystonia; TS, Tourette's syndrome; VA, ventral anterior nucleus thalamus; VIM, ventral intermediate nucleus thalamus; VOA/VOP, ventral oralis anterior/posterior nucleus thalamus. Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) flowchart showing the systematic identification and filtering of studies based on inclusion criteria. TABLE 2Dystonia subtypes included in studyDystonia subtypeNumber of patientsGeneralized142Segmental52Cervical317Craniocervical28Meige syndrome68Multisegmental2Multifocal4Focal17Orofacial3Truncal1DYT111DYT33DYT65DYT1122GNAL1GNAO‐11KMT2B5TWNK mutation2VPS163Tardive12Acquired23Dyskinetic cerebral palsy14Not specified126 Dystonia subtypes included in study Studies investigating dystonia combined with other non‐dystonia diagnoses were included (Table 2). Basal Ganglia LFPs in Dystonia Globus pallidus internus GPi recordings comprised the majority of lead targets in this review, consistent with the clinical treatment of dystonia. The GPi is a principal basal ganglia output nucleus and the predominant DBS target in dystonia. 25 The striatum connects to the GPI via the classical direct pathway, modulated primarily through D1 dopamine receptor activity. 26 Low‐frequency oscillatory activity Prominent low‐frequency activity (below 12 Hz) is a feature of idiopathic dystonia. 27 , 28 , 29 , 30 , 31 Silberstein et al found disease specific oscillatory activity in patients with dystonia and PD undergoing GPi DBS. LFP recordings were taken at rest, 2–7 days post‐operatively from externalized leads. 18 Three different states (off‐medication PD, on‐medication PD and dystonia) had sufficiently different patterns of LFP oscillatory activity allowing their identification. PD on‐medication had a relative increase in the activity in the 4–10 Hz band and less 11–30 Hz activity compared to the off medication state. Comparing off or on‐medication PD to dystonia, there was excessive low‐frequency (4–10 Hz) oscillations and reduced beta band activity in dystonia patients. Overall dystonia demonstrated the highest relative 4–10 Hz activity compared to the on‐ or off‐ medication PD state. Differences in GPi activity in dystonia and PD have been reported in other studies, with raised low‐frequency (below 12 Hz) a feature of dystonia and raised beta frequency (13–30 Hz) a consistent feature in PD. 32 , 33 Increased pallidal theta/low‐frequency oscillatory activity is present in idiopathic heterogenous dystonia cohorts. 34 , 35 , 36 , 37 , 38 , 39 CD and Meige syndrome showed similar spectral features. Zhu et al compared intra‐operative pallidal LFP recordings between CD and Huntington's disease. Abnormally increased low‐frequency oscillatory (6–14 Hz) activity was shown in CD but not Huntington's disease. 30 Zhang et al 40 recorded LFPs in 28 patients with Meige syndrome. Intra‐operative recordings were performed at rest from the GPi. Raised theta (4–8 Hz) oscillatory activity was present in Meige syndrome. The duration of theta burst activity was associated with pre‐operative symptom severity. GPi activity was symmetrical. GPi activity in inherited dystonias Inherited dystonia oscillatory activity varies by dystonia subtype. Prominent low‐frequency activity is present in DYT6 or DYT1. Several studies have investigated LFPs in myoclonus dystonia in isolation or as part of a heterogenous group of dystonias. Intra‐operative LFP recordings demonstrated raised 4–8 Hz activity in DYT11+ myoclonus dystonia and GD. Differences in GPi neuron bursting patterns were seen during the same recordings. 41 Idiopathic and DYT6 had raised theta and alpha frequency activity compared to myoclonus dystonia. 42 In contrast, DYT1 and DYT3 showed raised GPi beta activity. GPi recordings in one pediatric patient with DYT1 showed raised beta band activity. Using the Percept IPG in pediatric DYT1, two peaks in low beta (13.67 Hz) and high beta (25.39 Hz) were seen at initial programming. A desynchronization of beta band activity and clinical improvement were shown with stimulation. 43 Similarly in DYT3, raised beta activity may be present in acute or chronic recordings. Intra‐operative recordings in one patient showed increased beta activity at rest, with HFS suppressing increased alpha and beta oscillatory activity. 44 Donlon et al reported chronic recordings over 12 months in two DYT3 patients. These showed different spectral characteristics with no definitive peaks in one patient and raised beta activity in the other. Raised beta activity if present, may be suppressed with intra‐operative or chronic stimulation up to 12 months. 45 Therefore inherited dystonias do not appear to have a common oscillatory activity in contrast to idiopathic dystonia. GPi activity in acquired dystonia Acquired dystonias have variable oscillatory patterns in comparison to idiopathic dystonias. TD shows similar spectral characteristics to CD. Intra‐operative GPi recordings from seven cervical and five TD patients showed increased oscillatory activity over 1–7 Hz. Sensory palmar stimulation reduced the peak frequency in cervical but not TD, suggesting different modulation of abnormal oscillations despite similar baseline characteristics. 29 Acute post‐operative recordings 2–6 days post‐operatively in a group including TD showed raised low‐frequency oscillatory activity (5–12 Hz). Administration of a dopamine inhibitor produced an increase in low beta (13–20 Hz) activity. This suggests that beta activity is not specific to PD, but rather a marker of the depleted dopamine state seen in PD. 37 In comparison to idiopathic dystonia, raised beta activity is a feature in some acquired dystonia. 10 pediatric patients with medically refractive status dystonicus were implanted with GPi DBS and Percept IPG. During status dystonicus episodes raised beta oscillatory activity and burst amplitudes were seen compared to a non‐status dystonicus state. Beta activity was correlated with worse quality of life scores, leading the authors to propose excessive beta activity is a biomarker for status dystonicus. 46 Intra‐operative GPi recordings from four acquired dystonias showed prominent 13–35 Hz activity. 44 Outside the GPi, acquired dystonia demonstrates prominent oscillatory activity in other network nodes. Acute post‐operative STN recordings in chorea‐acanthocytosis dystonia showed coherence between STN and mesial sensorimotor cortex was shown in the 28–35 Hz frequency band bilaterally. Right STN was coherent with medial prefrontal cortex in the 4–12 Hz frequency. 47 Chronic recordings from dentate nucleus in three patients with dyskinetic cerebral palsy showed prominent alpha activity (9.8 Hz). HFS lead to a decrease in the variability of alpha oscillatory. 48 Thus heterogenous oscillatory activity in pallidal and other basal ganglia nodes is a feature of acquired dystonia. Excessive beta‐band activity may represent specific clinical states rather than a ubiquitous pathophysiological oscillatory signal in acquired dystonia. Low‐frequency activity and symptom severity Dystonia clinical severity is correlated with pallidal theta band activity. Neumann et al 21 included 27 patients with CD. LFPs were recorded 2–7 days post‐operatively via externalized leads with patients at rest. Two distinct peaks in the theta (4–12 Hz) and beta (13–35 Hz) were recorded. The peak power in the theta but not beta band was correlated with symptoms severity measured on TWSTRS. Acute post‐operative LFP recordings in a mixed primary dystonia cohort also demonstrated a correlation between GPi low‐frequency oscillatory activity and symptoms. 49 Chronic GPi LFP recordings in idiopathic and acquired pediatric dystonia demonstrate circadian modulation. Delta/theta and gamma power was increased during symptomatic dystonic periods. 50 Asymmetrical network dysfunction Pallidal activity is asymmetrical in CD and may be related to the direction of the abnormal head posture. Intra‐operative pallidal LFP oscillatory activity and single unit spontaneous discharge patterns altered between sides in patients with laterocollis and asymmetrical neck tone. The asymmetry correlated with the magnitude of laterocollis in this group. 51 Another study in CD reported the pallidum contralateral to the head turn showed more inhibition than the unilateral side and beta band activity (12–30 Hz) was seen in this group. 52 Lee et al recorded intra‐operative LFPs in 11 patients with CD. Raised low‐frequency activity (4–12 Hz) and beta activity (12–30 Hz) were recorded and were asymmetrical in torticollis or laterocollis independent of the side the head moved to. In contrast, retrocollis did not demonstrate GPi asymmetry. 53 Summary In idiopathic, inherited and acquired dystonia, GPi LFP recordings most consistently demonstrate abnormal low‐frequency oscillatory. The inconsistent presence of raised beta activity is likely a result in of the variability between dystonia subtypes, rather than a reliable biomarker. These findings suggest that asymmetric GPi activity represents broad network imbalance rather than purely lateralized motor output. Globus pallidus externus The GPe is part of the classical indirect pathway that is modulated by D2 dopamine receptors. 54 Compared to GPi recordings, there is a paucity of LFP recordings from the GPe and where present have been in addition to GPi recordings. Despite this, the available evidence does suggest that network abnormalities in dystonia are not limited to the GPi or basal ganglia output nodes but involve pallidal circuitry. Studies reporting simultaneous GPi and GPe LFP recordings show raised low‐frequency activity is present in both structures, highest in the theta and alpha bands. GPe theta activity may be lower in power relative to GPi activity, but its presence suggests that dysfunction is not limited solely to the GPi and a broader abnormal pallidal network may contribute to pathophysiology in dystonia. Interestingly there was no asymmetry present in GPe in contrast to GPi in CD. 28 Intra‐operative recordings in heterogenous idiopathic dystonia with simultaneous putamen, GPe and GPi recordings demonstrate raised low‐frequency activity. In contrast to GPi recordings, low‐frequency GPe or putamen activity did not correlate with symptom severity. 49 Although low‐frequency oscillations are observed in the GPe, their lack of correlation with symptom severity and the small number of studies limit their use as independent biomarkers. Putting GPe recordings into context, they support the presence of broad pallidal circuit dysfunction. Subthalamic nucleus Interest in the STN as a dystonia target is increasing. 55 However STN recordings show heterogenous oscillatory activity. These tend to depend on dystonia subtype and clinical state rather than a consistent STN dystonia signature. Guan‐Yu et al recorded acute post‐operative recordings from 17 CD, 19 Meige syndrome and 9 generalized dystonia (GD) patients and showed raised low‐frequency and high‐beta activity localized to the sensorimotor STN. Low‐frequency activity was correlated with symptoms in generalized and CD but not Meige syndrome. Oscillatory activity may be phenotype dependent. 56 STN LFP recordings were compared between dystonia and PD. STN 0–10 Hz power is higher in dystonia than PD, while dystonia shows reduced beta activity. Different oscillatory activity is present in different phenotypes. The low‐frequency activity in dystonia was not confined to the GPi, emphasizing the network dysfunction present in dystonia as opposed to a single nucleus issue. 38 In contrast, intra‐operative STN recordings showed increased beta activity in dystonia and PD with little difference in spectra between the two conditions. 57 Acute post‐operative STN recordings in cranio‐cervical dystonia did not show raised theta oscillatory activity at rest. STN HFS in this cohort produced broadband suppression, evoked resonant neural activity and finely tuned gamma oscillatory activity, similar to PD stimulation effects. 58 Thus STN oscillatory patterns in dystonia may share common features with PD, particularly during HFS, preventing the identification of each condition from the oscillatory pattern. Recordings from the STN in GD are rare. A single case report with acute post‐operative recordings at rest showed raised low‐frequency activity in the STN, consistent with the GPi. 59 From the above evidence, STN LFP recordings in dystonia show etiology and phenotype specific oscillatory activity. Low‐frequency activity correlated with some dystonia subtypes. The presence of raised low‐frequency activity beyond the pallidum in dystonia points to a network level disruption of signaling rather than pathology in a single basal ganglia nucleus. Thalamus Thalamic nuclei recordings have been performed in a limited number of studies, as part of multisite recordings simultaneously from the thalamus, GPi and STN. These recordings have given insights into the classic basal ganglia “rate model” of dystonia. During movement, thalamic nuclei (VOA/VOP and VP) oscillatory activity increased compared to rest. In addition, GPi and thalamic activity both increased during voluntary movement. 60 , 61 This contradicts the current model of decreased GPi activity leading to overactivity of the thalamus and subsequent dystonic muscle activity. GPi to thalamic inhibitory pathway is intact, but dystonic activity may be more accurately explained by abnormal patterns of activity, rather than activity levels. 61 Cortical and cortico‐basal ganglia interactions Task modulation of LFPs recorded from the GPi and STN have given insights into physiological processes during decision making and sensory stimulation. Navid et al performed a flanker task with LFPs recorded from the STN in PD and GPi in dystonia. Cue triggered theta activity increased in both groups. This theta activity likely represents cognitive processing via cortical to basal ganglia pathways. Response triggered beta activity was significantly higher in the GPi compared to the STN group. During a motor response, beta frequency may play an important role in signaling between cortex, STN and GPi networks. 62 GPi theta activity is increased during error response in the flanker task. Pallidal theta increases before cortical theta activity, with a flow in direction from the pallidum to cortex during conflict. This suggest pallidal output modulates cortical activity during decision making. 63 , 64 Sensory trick (ST) studies showed modulation of low‐frequency activity in CD. Effective STs were associated with reduced electromyography (EMG) activity and suppression of 6–8 Hz cortical and pallidal oscillations. In contrast, ineffective STs produced increases in the same band. EMG muscle activity was suppressed with an effective ST prior to facial contact. 65 Intra‐operative GPi recordings showed raised theta/delta activity at rest in CD and TD. Sensory palmar stimulation produced a reduction in peak frequency and power in CD patients compared to rest. This was not seen in recordings in TD under the same condition. This modulation of low theta/delta activity may be a sensory processing signature in CD. 29 Chronic Percept™ recordings suppressed 3–12 Hz activity during a ST in cervical, myoclonus and multifocal dystonia. During a proprioception task performed intra‐operatively, alpha band activity was increased in CD. 66 GPi alpha activity is modulated by neck vibration, with isolated CD showing a greater change in alpha activity than GD with cervical involvement. This alpha activity represents pallido‐cerebellar network coupling. Thus CD exhibits enhanced sensitivity of sensory network inputs than GD. 67 An interplay between basal ganglia structures and cortex is evident during information processing in dystonia. Suppression of increased muscle activity prior to facial contact during a ST points to complex network processing involving basal ganglia, cortex and muscle. The suppression of 6–8 Hz activity during the ST, in association with an improvement in symptoms suggests that modulation of abnormally increased pallidal low‐frequency activity is a therapeutic strategy to improve symptoms in dystonia. Network coupling Simultaneous pallidal, cortical and muscle recordings provide direct evidence that abnormal oscillatory activity in dystonia represents disordered network interactions rather than dysfunction of a single and isolated nucleus. GPi—muscle Low‐frequency pallidal activity is coherent with and drives dystonic EMG activity, supporting a link between basal ganglia output and abnormal muscle activitation. 68 , 69 Thirteen patients with dystonia and cervical involvement had acute post‐operative LFP recordings from GPi and simultaneous electromyography (EMG) from affected neck muscle. Coherence between GPi and muscle during involuntary muscle activity was observed in the 4–10 Hz frequency. 31 Similarly, raised theta band (4–8 Hz) GPi activity was coherent with muscle activity during myoclonus muscle activity. 70 Raised 3–15 Hz activity was suppressed during movement and was coherent with GPi and recordings from affected muscles, suggesting a relationship between abnormal pallidal oscillations and dystonic muscle activity. 71 These suggest low‐frequency pallidal activity is involved in dystonic muscle activity. However, evidence is lacking to say this is causative at present. GPi—cortex Recordings from multiple basal ganglia structures show how low‐frequency oscillations are present across pallidal circuitry. Lofredi et al implanted bilateral GPi DBS in a mixed group of 10 patients with dystonia. Simultaneous recordings from the putamen, globus pallidus externus (GPe) and GPi were possible with an eight contact lead. Low‐frequency oscillations were present in the putamen, GPe and GPi, with GPi low‐frequency oscillatory power significantly associated with dystonia severity. The link between low‐frequency power and symptoms was not replicated in the putamen or GPe based on the contact positions within these structures limited by anatomy and surgical approach. Coherence between low‐frequency activity in the putamen and GPi was linked to the severity of symptoms, but not between the putamen and GPe, suggesting diffuse basal ganglia network involvement. 49 Neumann et al 72 used pallidal LFPs and cortical whole head magnetoencephalography (MEG) to demonstrate network coupling in dystonia. The patient group consisted of idiopathic generalized, segmental and CD in adults. Coherence was present between pallidum and multiple cortical areas across multiple frequencies and spatially separate areas, representing network activity. Using Granger causality, the direction of information flow was from the GPi in the theta range (4–8 Hz) to temporal cortex, from cerebellum to GPi in the alpha range (7–13 Hz) and from motor and pre‐motor cortices to the GPi in the beta range. Pallido‐cerebellar alpha band coherence inversely correlated with TWSTRS severity, suggesting cerebellar to basal ganglia signaling may influence dystonia severity. 72 Abnormal low‐frequency oscillatory activity represents widespread network dysfunction involving basal ganglia, cortical and cerebellar circuitry. Modulation by stimulation Acute stimulation effects Acute HFS rapidly modulates pallidal oscillatory activity. However, effects differ across dystonia phenotypes. Barow et al 22 recorded acute post‐operative LFPs at rest, during and immediately after HFS in a phasic and tonic dystonia cohort. In phasic dystonia, HFS suppressed raised low‐frequency activity (4–12 Hz). Decreased coherence immediately following HFS between pallidum and cortex and muscle was also observed. These changes were not observed in tonic dystonia, suggesting a phenotype specific network response to stimulation. Piña‐Fuentes et al recorded LFPs from the GPi of a mixed dystonia cohort in stimulation naïve and chronic stimulation patients during lead insertion and battery changes respectively. Low‐frequency oscillatory activity was raised in dystonia and mild suppression was seen rapidly with continuous stimulation that did not correlate with symptom improvement. Stimulation induced suppression of low‐frequency activity may be gradual and remain suppressed after stimulation cessation. Clinical and electrophysiological changes may occur over different time‐scales. 73 Stimulation effects in inherited dystonia are even more limited. Giorni et al recorded intra‐operative LFPs in DYT6 dystonia revision surgery four years following insertion. Compared to stimulation naïve DYT6, HFS suppressed gamma band activity compared to no stimulation. 74 Multisite recordings from thalamus, STN and GPi in pediatric and young adult dystonia reported stimulation travels along physiological pathways more frequently in an orthodromic direction. GPi stimulation may modulate information flow through pallido‐thalamic circuitry. 75 Overall, acute stimulation modulates pallidal low‐frequency activity, but its relationship to clinical symptoms varies and studies remain scarce. Chronic stimulation effects Sensing enabled IPGs have made chronic LFP recordings off or on stimulation possible. Oscillatory activity can change over longer time scales and the relationship to clinical symptoms can vary. Chronic stimulation effects following a period of effective stimulation over months were investigated in a heterogenous dystonia group of nine patients. Stimulation was ceased and pallidal LFP recordings were combined with symptom severity over five time points over 7 h. Pallidal low‐frequency LFP power re‐emerged off stimulation and was correlated with increased symptom severity as the time from stimulation cessation increased. 76 Chronic pallidal recordings from sensing‐capable IPGs have been reported in one patient with CD. LFPs were recorded over a 27 month period. Raised low‐frequency (3–12 Hz) and gamma (60 Hz) were present off stimulation. Stimulation suppressed these two bands and induced an increase in gamma oscillatory activity circa 65 Hz and was associated with clinical improvement and reduction in dystonia severity. 77 Ledingham et al recorded LFPs chronically in a patient with CD. Theta and beta peaks were recorded, with the stimulation contact chosen that contained the highest theta power. Stimulation suppressed low‐frequency power and was associated with an improvement in dystonic symptoms. 78 Chronic recordings from three patients with GD investigated the effects of overstimulation on symptoms and LFPs. Overstimulation was defined as worsening of clinical symptoms during an actively monitored session. Overstimulation lead to a simultaneous worsening of dystonia and an increase in broadband oscillatory activity across 5–55 Hz. Network dysfunction may occur at insufficient or excessive stimulation. 79 In inherited dystonia, pallidal LFPs in DYT3 dystonia and their response to stimulation have been recorded. Raised beta activity prior to stimulation was present in one patient and was suppressed with HFS over a 12 month period that correlated with an improvement in clinical severity, although further evidence in genetic dystonia is lacking. 45 The effects of pallidal HFS are not well characterized. Few studies report suppression of abnormal low‐frequency oscillatory activity in certain dystonia phenotypes, in particular idiopathic cervical and phasic dystonia phenotypes. The changes in oscillatory activity with stimulation vary and at present do not predict clinical response. The current evidence suggests that stimulation exerts its effects by modulating network activity, rather than a single frequency pathological oscillatory activity. LFPs as biomarkers and adaptive DBS in dystonia Beta frequency activity is a reliable biomarker of disease state in PD. This biomarker is reliably present in over 90% of STN and correlates with symptom severity. 80 , 81 Suppression of beta activity by dopamine replacement or HFS reliably correlates with less symptom severity. 11 , 13 The reliability of beta as a biomarker has paved the way for adaptive DBS (aDBS) paradigms based on beta activity in PD. 82 In contrast, reliable biomarkers in dystonia are lacking. An adaptive stimulation protocol in dystonia was trialed from pallidal recordings in idiopathic dystonia with low‐frequency amplitude bursts as the biomarker to correspond to symptoms. This paradigm did not produce acute changes in low‐frequency oscillatory power or symptoms. Conventional continuous DBS did show mild suppression of low‐frequency activity, although this was independent of an improvement in symptoms. Low‐frequency activity may not represent a real‐time marker of dystonia disease state. 73 Motor cortical theta bursts were trialed as a biomarker in one patient with idiopathic CD implanted with STN leads in a proof‐of‐principle study. Off stimulation increased theta activity was present throughout the cortico‐subthalamic circuit. Suppression of theta activity and reduction in duration, rate and height of theta bursts were seen with stimulation. Larger peak amplitudes without increased dyskinesias were achieved with the aDBS protocol. Blinded clinical ratings were improved with aDBS versus continuous DBS. 83 LFP guided programming may be possible, but currently remains exploratory. Ledingham et al reported their experience in programming one CD patient based on LFP analysis. 78 Stimulation produced a decrease in low‐frequency power and an increase in beta power with an accompanying clinical improvement in symptoms. Similarly, Fasano et al 84 and Kelbert et al 43 used chronic LFP analysis in pediatric DYT1 dystonia to guide programming and response to stimulation. A clinical improvement was accompanied by a decrease in LFP power, most noticeably in the beta band. 43 To date, only a small number of proof‐of‐principle studies have been trialed for aDBS in dystonia, in contrast to PD, where aDBS paradigms are in routine clinical use in some centers. 82 Discussion This systematic review shows low‐frequency oscillatory activity is the most consistent electrophysiological feature across dystonia subtypes and basal ganglia nuclei. This abnormal activity is not limited to a single basal ganglia nucleus and is expressed across different phenotypes, clinical states and recordings conditions. Low‐frequency activity is present across the GPi, GPe, putamen and STN, is coherent with cortical and muscle activity and is modulated by sensory input and stimulation. 27 , 28 , 29 , 30 , 31 , 49 , 56 , 65 These findings suggest that dystonia is characterized by a disordered network activity rather than dysfunction of an isolated nucleus or single pathological frequency (Figure 2). Summary of local field potential findings in dystonia. (A) Network model of oscillatory dysfunction in dystonia (hyperdirect pathway from cortex to STN not visualized). (B) Clinical correlates of LFP findings. (C) Stimulation effects on oscillatory dysfunction in dystonia. (D) Summary. CD, cervical dystonia; DBS, deep brain stimulation; GPe, globus pallidus externus; GPi, globus pallidus internus; HFS, high‐frequency stimulation; LF, low‐frequency (3–12 Hz); LFP, local field potential; STN, subthalamic nucleus; TWSTRS, Toronto Western Spasmodic Torticollis Rating Scale. Conceptual framework of pathophysiology in dystonia While the studies included are heterogenous in methods and dystonia subtypes, several converging themes are present. Based on included studies, we propose a model where pathologic low‐frequency oscillatory activity represents abnormal interactions between the basal ganglia, cortical and cerebellar circuits. Central to this model is the GPi, one of the principle output nuclei of the basal ganglia. The association of raised low‐frequency activity with symptom severity, and its coherence with dystonic muscle activity, suggests this increased low‐frequency oscillatory activity potentially reflects pathological basal ganglia output that results in abnormal and hyperkinetic motor patterns. 21 , 68 , 69 However, these abnormalities are not confined to the GPi. This low‐frequency activity is present throughout multiple basal ganglia nuclei including the GPe, STN and putamen, rather than a single nucleus. 49 , 56 Dystonia activity may therefore not be driven be a single area of pathology, but as a result of widespread network dysfunction. Coherence between pallidal, cortical and muscle activity promotes the idea that this abnormal signal spreads throughout motor circuits and produces dysfunction. 31 , 72 Cerebellar networks have increasingly been recognized as playing an important role in the framework for dystonia network dysfunction. Impaired cerebellar‐basal ganglia connectivity is present in dystonia. 10 Pallido‐cerebellar coupling, mainly in the alpha range, and the relationship between pallido‐cerebellar connectivity and symptom severity, suggest that cerebellar pathways may modulate or attempt to compensate for the abnormal motor outputs from the pallidum in dystonia or may be necessary for normal motor control that is impaired in dystonia. 72 GPi DBS may restore network connectivity patterns in dystonia closer to that of healthy controls. 85 This fits in with proposed models of dystonia as a disorder of impaired integration across basal ganglia‐cerebellar‐cortical pathways. Within this network, low‐frequency oscillatory activity may represent a pathological state of excessive synchronization leading to impaired motor control and sensorimotor integration. Higher frequency activity, including gamma oscillations have also been demonstrated and may reflect pro‐kinetic signals, similar to entrained gamma oscillations following STN DBS in PD. 50 , 74 , 86 However, the significance of gamma frequency oscillatory activity in dystonia is not well understood. The effects of DBS further promote a network dysfunction. Stimulation may suppress low‐frequency oscillatory activity and modulate coherence between brain regions in some forms of dystonia, but this suppression is not reliably linked to symptom improvement. 22 , 73 Therapeutic effects of stimulation may therefore be as a result of a network reorganization rather than suppression of a single frequency's pathological oscillatory activity. From the evidence above, this framework supports the hypothesis that dystonia develops as a result of distributed network dysfunction, with low‐frequency oscillatory activity arising in the basal ganglia and spreading throughout the cortex and cerebellum, to produce the hyperkinetic dystonia phenotype. Implications for clinical heterogeneity This model may explain the clinical heterogeneity in dystonia. The wide variation in oscillatory activity across idiopathic, inherited and acquired dystonia subtypes is unsurprising considering the wide variation in phenotypes and does not represent a single disease specific oscillatory signature. 27 , 28 , 29 , 30 , 31 , 41 , 42 , 43 , 44 , 47 The difference in electrophysiological signatures may reflect differences in how network dysfunction is distributed across basal ganglia‐cortico‐cerebellar circuits. This could explain why low frequency oscillatory activity is not consistently expressed across idiopathic, inherited and acquired dystonia and why different nuclei show different involvement based on phenotype and clinical state. Implications for stimulation effects The variable effects of stimulation provide further evidence of a network‐based mechanism of DBS action, with stimulation acting on different circuits rather than a single pathological oscillatory activity. Low‐frequency activity may be readily suppressed in certain dystonia subtypes and coupling between pallidum, cortex and muscle may be modulated, but symptomatic improvement may not be seen in tandem with suppression of low‐frequency activity. 22 , 73 Therapeutic effects of stimulation may occur over a longer time. Stimulation benefits may result as a cumulative network organization over time, rather than a rapid suppression of abnormal low‐frequency activity. In addition, different dystonic phenotypes also respond differently to stimulation, such as phasic and tonic phenotypes. Different network abnormalities may represent specific phenotypes and stimulation effects may therefore differ depending on the network abnormality with the aim to restore the network to a more physiological state. 22 Implications for biomarkers and adaptive DBS In contrast to beta activity in PD, a reliable biomarker in dystonia has not been identified. Low‐frequency activity is inconsistently expressed, does not reliably represent the disease state and is not reliably modulated in response to stimulation. Therefore using a single frequency like beta activity in PD as a basis for adaptive DBS in dystonia is unlikely to work. Potential adaptive paradigms in dystonia may require biomarkers that include network coherence, temporal dynamics or non‐oscillatory recordings to better assess the clinical state. Current aDBS efforts are exploratory and experimental. The longer time to response with stimulation in dystonia further complicates matters, with current adaptive DBS responses to biomarker change over seconds to minutes. Longer‐time biomarkers of disease status may need to be identified before adaptive stimulation strategies are trialed in dystonia. Limitations The studies included in the review are at high risk of bias. Most studies included were small cohorts of a heterogenous group of dystonias with variations in recording conditions and analysis. Acute and chronic recordings reflect different physiological conditions and their direct comparison may be limited. While larger groups of subtypes of dystonia are desirable, the studies reflect a typical movement disorder clinic dystonia cohort and the dystonia LFP literature. Some dystonia subtypes have been reported in low numbers and are not well represented in the review. LFP findings in common types of dystonia may not be generalizable to more rare subtypes. Chronic LFPs are limited to case reports, but the widespread availability of sensing‐capable IPGs offers a unique opportunity in future. Conclusions Local field potentials in dystonia demonstrate widespread network abnormalities in network synchronization in the basal ganglia, cortical and cerebellar circuits. Oscillatory activity in dystonia represents widespread network dysfunction that is not confined to a single basal ganglia nucleus. This explains the heterogeneity in phenotypes and response to stimulation in dystonia. The availability of sensing‐capable IPGs provide a unique opportunity in the study of network dysfunction in dystonia. Further insights into network abnormalities may aid in the treatment of dystonia. Ethics Statement (1) This paper is a systematic review of previously published articles and so ethical approval by the institutional ethics review board was not required. (2) Consent was not required due to being a systematic review of previously published articles. (3) 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. Author Roles (1) Research project: A. Conception, B. Organization, C. Execution; (2) Analysis: A. Design, B. Execution, C. Review and Critique; (3) Manuscript Preparation: A. Writing of the first draft, B. Review and Critique. J.H.: 1A, 1B, 1C, 2A, 2B, 2C, 3A, 3B E.D.: 1A, 1B, 2A, 2B, 2C, 3A, 3B AG: 1B, 2A, 2B, 2C, 3A, 3B RM: 2C, 3B C.F.: 2C, 3B R.A.W.: 1B, 2A, 2B, 2C, 3A, 3B C.M.: 1A, 1B, 1C, 2A, 2B, 2C, 3A, 3B Disclosures JH, ED, CM: educational grant from Medtronic and Boston Scientific. CF: Grants from Michael J. Fox Foundation and Abbvie. Support for attending meetings from Ipsen, Abbvie, Clonmel Healthcare. Advisory board for Teva. RW: educational grant from Medtronic for unrelated fellowship funding. AG, RM: no financial disclosures. Funding This research was supported by Strategic Academic Recruitment (StAR) MD Programme grant from Royal College of Surgeons in Ireland (24225A010). Financial Disclosures and Conflicts of Interest Author disclosures are available in the Supporting Information. Supporting information TABLE S1. ROBINS‐E (Risk Of Bias in Non‐Randomized Studies—of Exposures). Supporting Information: COI_disclosure Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request. References Associated Data Supplementary Materials TABLE S1. ROBINS‐E (Risk Of Bias in Non‐Randomized Studies—of Exposures). Supporting Information: COI_disclosure Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request.