Preliminary clinical and electrophysiological findings of NAc/ALIC deep brain stimulation strategy for treatment-resistant depression Authors: Wei Zhang, Botao Xiong, Yukun Kang, Yuan Gao, Ruiqing Yang, Qiang Wang, Wei Wang Published: 2026 ABSTRACT SummaryEffective therapeutic options for treatment-resistant depression remain limited. This study investigated the clinical and electrophysiological effects of dual-target deep brain stimulation of the nucleus accumbens and anterior limb of the internal capsule. In a small patient cohort, the intervention was associated with reduced depressive symptoms and improved cognitive performance. Concurrent electrophysiological recordings revealed that clinical improvement correlated with elevated gamma oscillations in the stimulated target and increased theta-band power in frontal-limbic regions. These integrated findings provide a preliminary framework for understanding therapeutic mechanisms and suggest potential biomarkers for optimizing deep brain stimulation therapy in treatment-resistant depression. Introduction Depression is a widespread mental health disorder that significantly impairs social psychological functioning and quality of life, affecting over 280 million individuals globally.1 It is characterized by persistent low mood, loss of interest or pleasure, and is frequently accompanied by sleep disturbances and cognitive deficits.1,2,3,4 Current treatment options include pharmacotherapy,5 psychotherapy,6,7,8,9 transcranial magnetic stimulation,10 electroconvulsive therapy,11phytochemicals,12,13 alternative approaches including ozone therapy,14,15 etc. Despite these interventions, approximately 20%–30% of patients with depression respond poorly to conventional treatments, known as treatment-resistant depression (TRD).16,17,18 Deep brain stimulation (DBS) has emerged as a promising therapeutic approach for TRD. In 2005, Mayberg et al. published the first study on DBS targeting the subcallosal cingulate (SCC) in six patients with TRD, four of whom met response criteria—defined as a reduction of more than 50% in Hamilton depression rating scale (HAMD) scores.19 Subsequent trials have explored different brain targets, with estimated response, remission, and relapse rates of 56%, 35%, and 14%, underscoring the therapeutic potential of DBS.20 Three targets have been studied in large randomized controlled trials (RCTs): SCC,21,22,23 ventral capsule/ventral striatum (VC/VS),24 ventral anterior limb of internal capsule (vALIC).25 Among these, only vALIC trials showed significant variability, though the response rate reached 40%. Nucleus accumbens (NAc) plays a key role in emotional regulation and cognitive processing. It receives dopaminergic projections from multiple brain regions and relays signals to other structures, thereby modulating widespread neural circuitry.26,27 Given its anatomical proximity to the ALIC, simultaneous stimulation of both NAc and ALIC is feasible. Initially, the dual-target DBS (NAc-ALIC) was applied in refractory obsessive-compulsive disorder, during which notable improvements in comorbid depressive symptoms were observed.28,29,30 However, only a limited number of studies have investigated dual-target DBS specifically for TRD.31,32 Therefore, this study aimed to evaluate the clinical efficacy of dual-target DBS in patients with TRD. Additionally, we conducted long-term recordings of local field potentials (LFPs) and magnetoencephalography (MEG) to identify associated electrophysiological features, which may provide insights for optimizing future DBS therapy. Results Demographic and clinical characteristics Between June 2022 and January 2024, seven patients diagnosed with TRD were enrolled and underwent DBS surgery in the neurosurgery department. All patients completed at least six months of follow-up. The cohort consisted of two males and five females, with a mean age of 25.71 ± 8.18 years and a mean illness duration of 6.14 ± 2.67 years. Notably, one patient had previously undergone vagus nerve stimulation three years earlier but subsequently had the electrodes and pulse generator removed due to insufficient therapeutic effect. Nevertheless, this patient still achieved a greater than 50% reduction in HAMD-24 score at one year after DBS implantation. (Table 1).Table 1Demographic and clinical informationVariableTRD (n = 7) Mean ± SD (range)PGender (male/female)2/5–Age (years)25.71 ± 8.18 (17–36)–Disease duration (years)6.14 ± 2.67 (3–10)–Suicide behavior (yes/no)3/4–Past TMS (yes/no)7/0–Past ECT (yes/no)6/1–Past psychotherapy (yes/no)7/0–Past ketamine(yes/no)6/1–Past psychiatric surgery (yes/no)1/6–HAMD-24Baseline38.14±13.68 (24-65)six months after DBS16.14±7.66 (5-29)0.00795%CI14.98,33.59HAMABaseline26.29±5.37 (20-35)six months after DBS12.43±5.85 (3-22)0.00295%CI7.37,27.77ECT, electroconvulsive therapy; HAMD-24, 24-item Hamilton depression rating scale; HAMA, Hamilton anxiety rating scale; TMS, transcranial magnetic stimulation; TRD, treatment-resistant depression. Table 1Demographic and clinical informationVariableTRD (n = 7) Mean ± SD (range)PGender (male/female)2/5–Age (years)25.71 ± 8.18 (17–36)–Disease duration (years)6.14 ± 2.67 (3–10)–Suicide behavior (yes/no)3/4–Past TMS (yes/no)7/0–Past ECT (yes/no)6/1–Past psychotherapy (yes/no)7/0–Past ketamine(yes/no)6/1–Past psychiatric surgery (yes/no)1/6–HAMD-24Baseline38.14±13.68 (24-65)six months after DBS16.14±7.66 (5-29)0.00795%CI14.98,33.59HAMABaseline26.29±5.37 (20-35)six months after DBS12.43±5.85 (3-22)0.00295%CI7.37,27.77ECT, electroconvulsive therapy; HAMD-24, 24-item Hamilton depression rating scale; HAMA, Hamilton anxiety rating scale; TMS, transcranial magnetic stimulation; TRD, treatment-resistant depression. Demographic and clinical information ECT, electroconvulsive therapy; HAMD-24, 24-item Hamilton depression rating scale; HAMA, Hamilton anxiety rating scale; TMS, transcranial magnetic stimulation; TRD, treatment-resistant depression. Primary outcomes The HAMD-24 scores decreased significantly from 38.14 ± 13.68 at baseline to 16.14 ± 7.66 at the six-month follow-up (paired t test, p = 0.007). Five patients (71.4%) met the response criterion (≥50% reduction in HAMD-24 score), two of whom reached clinical remission (HAMD-24 score <8). The remaining two patients (28.6%) showed a partial response (Table 1; Figure 1A). Among the four patients followed for more than one year, three maintained a ≥50% reduction in HAMD-24 score. One patient experienced depressive relapse following depletion of the implantable pulse generator battery. Of the two patients followed for over two years, one continued to meet response criteria, while the other relapsed due to reluctance to adjust stimulation parameters.Figure 1Clinical outcomes following DBS(A) HAMD-24 scores of baseline, one month, three months, and six months after DBS (paired t test, n = 7).(B) HAMA scores of baseline, one month, three months, and six months after DBS (paired t test, n = 7).(C) MCCB scores of baseline, and six months after DBS (paired t test, ∗p < 0.05, n = 5).(D) The incidence of physical symptoms presenting in TRD patients. DBS, Deep brain stimulation; HAMD-24, 24-item Hamilton depression rating scale; HAMA, Hamilton anxiety rating scale; MCCB, MATRICS consensus cognitive battery; TRD, treatment-resistant depression. Clinical outcomes following DBS (A) HAMD-24 scores of baseline, one month, three months, and six months after DBS (paired t test, n = 7). (B) HAMA scores of baseline, one month, three months, and six months after DBS (paired t test, n = 7). (C) MCCB scores of baseline, and six months after DBS (paired t test, ∗p < 0.05, n = 5). (D) The incidence of physical symptoms presenting in TRD patients. DBS, Deep brain stimulation; HAMD-24, 24-item Hamilton depression rating scale; HAMA, Hamilton anxiety rating scale; MCCB, MATRICS consensus cognitive battery; TRD, treatment-resistant depression. Secondary outcomes At the six-month follow-up, the Hamilton anxiety rating scale (HAMA) score decreased significantly from 26.29 ± 5.37 to 12.43 ± 5.85 (paired t test, p = 0.002) (Table 1; Figure 1B). Performance on the trail making test also improved, as reflected by a significant reduction in completion time (p = 0.04), indicating enhanced psychomotor speed. Significant improvements were observed in visual memory (paired t test, p = 0.04) and category fluency (paired t test, p = 0.02) in MATRICS consensus cognitive battery (MCCB). However, other MCCB subtests showed no statistically significant change (paired t test, p > 0.05) (Table 2; Figure 1C). No significant differences were found before and after DBS in scores on the dimensional anhedonia rating scale (DARS), temporal experience of pleasure scale (TEPS), World Health Organization quality of life (WHOQOL), global assessment of function (GAF), and Pittsburgh sleep quality index (PSQI) at the assessed time point (paired t test, p > 0.05).Table 2Cognitive function outcomes during DBSVariableBaseline6 monthsT95%CIPTMT(s)53.19 ± 17.3533.63 ± 8.912.961.26, 37.850.04SC46.00 ± 19.8554.80 ± 13.83−1.12−62.44, 26.440.32HVLT-R21.80 ± 4.3224.20 ± 8.98−0.78−10.93, 6.130.47DS20.80 ± 3.4924.20 ± 4.38−1.61−9.26, 2.460.18Mazes test13.20 ± 3.8917.60 ± 5.03−1.61−11.98, 3.180.18BVMT-R22.78 ± 12.3139.07 ± 8.90−2.99−31.36, −1.200.04CF17.00 ± 7.5223.20 ± 7.39−4.23−10.26, −4.130.01MSCEIT56.04 ± 35.7979.63 ± 3.58−2.38−99.58, 7.570.07CPT2.11 ± 0.662.56 ± 0.32−2.55−0.93, 0.030.06BVMT-R, brief visuospatial memory test-revised; CF, category fluency; CPT, continuous performance test; DBS, deep brain stimulation; DS, digit span; HVLT-R, Hopkins verbal learning test-revised; MSCEIT, Mayer-Salovey-Caruso emotional intelligence test; SC, sign coding; TMT, trail making test. Table 2Cognitive function outcomes during DBSVariableBaseline6 monthsT95%CIPTMT(s)53.19 ± 17.3533.63 ± 8.912.961.26, 37.850.04SC46.00 ± 19.8554.80 ± 13.83−1.12−62.44, 26.440.32HVLT-R21.80 ± 4.3224.20 ± 8.98−0.78−10.93, 6.130.47DS20.80 ± 3.4924.20 ± 4.38−1.61−9.26, 2.460.18Mazes test13.20 ± 3.8917.60 ± 5.03−1.61−11.98, 3.180.18BVMT-R22.78 ± 12.3139.07 ± 8.90−2.99−31.36, −1.200.04CF17.00 ± 7.5223.20 ± 7.39−4.23−10.26, −4.130.01MSCEIT56.04 ± 35.7979.63 ± 3.58−2.38−99.58, 7.570.07CPT2.11 ± 0.662.56 ± 0.32−2.55−0.93, 0.030.06BVMT-R, brief visuospatial memory test-revised; CF, category fluency; CPT, continuous performance test; DBS, deep brain stimulation; DS, digit span; HVLT-R, Hopkins verbal learning test-revised; MSCEIT, Mayer-Salovey-Caruso emotional intelligence test; SC, sign coding; TMT, trail making test. Cognitive function outcomes during DBS BVMT-R, brief visuospatial memory test-revised; CF, category fluency; CPT, continuous performance test; DBS, deep brain stimulation; DS, digit span; HVLT-R, Hopkins verbal learning test-revised; MSCEIT, Mayer-Salovey-Caruso emotional intelligence test; SC, sign coding; TMT, trail making test. Gamma oscillations in NAc-ALIC showed a significant negative correlation with depression severity scores A total of 26 LFP recordings were obtained from five patients. Recordings before and six months after DBS were compared, revealing no statistically significant differences in LFP power at any contact point (Paired t test, p > 0.05, Figure S1). The other two patients could not undergo LFP recording due to unavailability of recording functionality in the implanted device. Spearman’s correlation analysis was performed to assess the relationship between relative PSD power and HAMD-24 scores. Results indicated a significant negative correlation between gamma-band power and HAMD-24 scores. Specifically, greater high-gamma power in the left NAc was associated with lower HAMD-24 scores (r = −0.41, p = 0.037, Figure 2C). Similarly, high-gamma power in the right ALIC correlated negatively with HAMD-24 scores (r = −0.55, p = 0.004, Figure 2D), as did low-gamma power in the right ALIC (r = −0.41, p = 0.038, Figure 2E) and high-gamma power in the right NAc–ALIC junction (r = −0.41, p = 0.037, Figure 2F).Figure 2Analysis pipeline and outcomes of local field potentials and magnetoencephalography(A and B) Local field potentials and magnetoencephalography analysis pipeline.(C) Correlation coefficient of high gamma between left NAc and HAMD-24 scores (Spearman’s correlation test, r = −0.41, p = 0.037, n = 26).(D) Correlation coefficient of high gamma between right ALIC and HAMD-24 scores (Spearman’s correlation test, r = −0.55, p = 0.004, n = 26).(E) Correlation coefficient of low gamma between right ALIC and HAMD-24 scores (Spearman’s correlation test, r = −0.41, p = 0.038, n = 26).(F) Correlation coefficient of high gamma between the junction of right NAc-ALIC and HAMD-24 score (Spearman’s correlation test, r = −0.41, p = 0.037, n = 26).(G–L) Relative theta power (5–8 Hz) in the red regions was significantly increased in left bankssts, left caudal anterior cingulate gyrus, left entorhinal cortex, left fusiform gyrus, left lateral orbitofrontal gyrus, right lingual gyrus, left medial orbitofrontal gyrus, right medial orbitofrontal gyrus, left para hippocampal gyrus, left pars orbitalis, left precuneus, and left temporal pole from different view.(M) Bar chart of increased relative theta power in twelve brain regions (paired t test, ∗p < 0.05, ∗∗p < 0.01, n = 4). ALIC, anterior limb of internal capsule; HAMD-24, 24-item Hamilton depression rating scale; NAc, nucleus accumbens; PSD, power spectrum density. Analysis pipeline and outcomes of local field potentials and magnetoencephalography (A and B) Local field potentials and magnetoencephalography analysis pipeline. (C) Correlation coefficient of high gamma between left NAc and HAMD-24 scores (Spearman’s correlation test, r = −0.41, p = 0.037, n = 26). (D) Correlation coefficient of high gamma between right ALIC and HAMD-24 scores (Spearman’s correlation test, r = −0.55, p = 0.004, n = 26). (E) Correlation coefficient of low gamma between right ALIC and HAMD-24 scores (Spearman’s correlation test, r = −0.41, p = 0.038, n = 26). (F) Correlation coefficient of high gamma between the junction of right NAc-ALIC and HAMD-24 score (Spearman’s correlation test, r = −0.41, p = 0.037, n = 26). (G–L) Relative theta power (5–8 Hz) in the red regions was significantly increased in left bankssts, left caudal anterior cingulate gyrus, left entorhinal cortex, left fusiform gyrus, left lateral orbitofrontal gyrus, right lingual gyrus, left medial orbitofrontal gyrus, right medial orbitofrontal gyrus, left para hippocampal gyrus, left pars orbitalis, left precuneus, and left temporal pole from different view. (M) Bar chart of increased relative theta power in twelve brain regions (paired t test, ∗p < 0.05, ∗∗p < 0.01, n = 4). ALIC, anterior limb of internal capsule; HAMD-24, 24-item Hamilton depression rating scale; NAc, nucleus accumbens; PSD, power spectrum density. Modulation of theta-specific networks was mediated by NAc-ALIC-DBS PSD was computed for 68 cortical regions following MEG acquisition. PSD values were normalized to relative power (range: 0–1), reflecting the proportional contribution of each of the six frequency bands to the total power. Following six months of DBS, an increase in theta frequency band was found in twelve distinct brain regions (paired t test): left bankssts (p = 0.015), left caudal anterior cingulate gyrus (p = 0.020), left entorhinal cortex (p = 0.012), left fusiform gyrus (p = 0.007), left lateral orbitofrontal gyrus (p = 0.002), right lingual gyrus (p = 0.036), left medial orbitofrontal gyrus (p = 0.003), right medial orbitofrontal gyrus (p = 0.013), left para hippocampal gyrus (p = 0.043), left pars orbitalis (p = 0.029), left precuneus (p = 0.021), and left temporal pole (p = 0.004). No statistically significant differences (paired t test, p > 0.05) were found in the other frequency bands and brain regions before and six months after DBS (Figures 2G–2M; Table 3). There is no significant statistical difference in granger causality connectivity before and six months after DBS (Wilcoxon signed-rank test, p > 0.05, Figure S2).Table 3PSD changes in MEG during DBSLocationBaseline6 monthsT95%CIPleft bankssts0.18 ± 0.030.25 ± 0.03−5.02−0.10, −0.020.015left caudal anterior cingulate gyrus0.19 ± 0.030.26 ± 0.02−4.50−0.11, −0.020.020left entorhinal cortex0.17 ± 0.030.27 ± 0.05−5.35−0.15, −0.030.012left fusiform gyrus0.18 ± 0.030.25 ± 0.03−6.34−0.09, −0.030.007left lateral orbitofrontal gyrus0.16 ± 0.020.25 ± 0.02−11.46−0.11, −0.060.002right lingual gyrus0.18 ± 0.030.25 ± 0.02−3.62−0.14, −0.010.036left medial orbitofrontal gyrus0.16 ± 0.020.23 ± 0.04−8.78−0.09, −0.040.003right medial orbitofrontal gyrus0.17 ± 0.020.23 ± 0.04−5.25−0.10, −0.020.013left parahippocampal gyrus0.18 ± 0.050.27 ± 0.02−3.38−0.17, −0.010.043left pars orbitalis0.15 ± 0.010.25 ± 0.06−3.94−0.19, −0.020.029left precuneus0.18 ± 0.040.24 ± 0.04−4.17−0.11, −0.010.021left temporal pole0.15 ± 0.020.26 ± 0.047.59−0.15, −0.060.004DBS, deep brain stimulation; MEG, magnetoencephalography; PSD, power spectrum density. Table 3PSD changes in MEG during DBSLocationBaseline6 monthsT95%CIPleft bankssts0.18 ± 0.030.25 ± 0.03−5.02−0.10, −0.020.015left caudal anterior cingulate gyrus0.19 ± 0.030.26 ± 0.02−4.50−0.11, −0.020.020left entorhinal cortex0.17 ± 0.030.27 ± 0.05−5.35−0.15, −0.030.012left fusiform gyrus0.18 ± 0.030.25 ± 0.03−6.34−0.09, −0.030.007left lateral orbitofrontal gyrus0.16 ± 0.020.25 ± 0.02−11.46−0.11, −0.060.002right lingual gyrus0.18 ± 0.030.25 ± 0.02−3.62−0.14, −0.010.036left medial orbitofrontal gyrus0.16 ± 0.020.23 ± 0.04−8.78−0.09, −0.040.003right medial orbitofrontal gyrus0.17 ± 0.020.23 ± 0.04−5.25−0.10, −0.020.013left parahippocampal gyrus0.18 ± 0.050.27 ± 0.02−3.38−0.17, −0.010.043left pars orbitalis0.15 ± 0.010.25 ± 0.06−3.94−0.19, −0.020.029left precuneus0.18 ± 0.040.24 ± 0.04−4.17−0.11, −0.010.021left temporal pole0.15 ± 0.020.26 ± 0.047.59−0.15, −0.060.004DBS, deep brain stimulation; MEG, magnetoencephalography; PSD, power spectrum density. PSD changes in MEG during DBS DBS, deep brain stimulation; MEG, magnetoencephalography; PSD, power spectrum density. Adverse effects Adverse effects related to the surgical procedure were not observed. All side effects associated with DBS therapy were transient and could be promptly controlled by parameter adjustments, with no permanent adverse outcomes for any patient. One patient experienced seizures immediately after DBS activation, which was attributed to be the incomplete resolution of perisurgical edema. Another exhibited symptoms resembling sympathetic hyperactivity during LFPs recording with stimulation turned off, which resolved once stimulation was resumed. Other side effects, including mild mania, dizziness, nausea, vomiting, and changes in appetite, were all alleviated following parameter modification. Discussion This study introduced an innovative approach by utilizing long pulse width, high frequency, and dual-target DBS (Figure 3) for TRD in the clinical research. The complete response rate was observed to be 71.4% at six-months follow-up, preliminarily indicating the efficacy of this therapy. Additionally, LFPs analysis have revealed an inverse correlation between gamma oscillations and HAMD scores. MEG analysis showed that DBS treatment increased theta oscillations in the left orbitofrontal cortex, bilateral medial frontal cortex, left limbic system, left parietal cortex, and right occipital cortex. These findings suggest that dual-target DBS is associated with distinctive electrophysiological alterations in both the nuclei and cortical regions.Figure 3Schematic diagram and 3D figure of dual-target deep brain stimulation(A) Atlas plate illustrating the planned electrode trajectory passing through the NAc and ALIC.(B) Postoperative three-dimensional reconstruction of implanted electrodes in seven patients, showing trajectories traversing both the NAc and ALIC. ALIC, anterior limb of internal capsule; NAc, nucleus accumbens. Schematic diagram and 3D figure of dual-target deep brain stimulation (A) Atlas plate illustrating the planned electrode trajectory passing through the NAc and ALIC. (B) Postoperative three-dimensional reconstruction of implanted electrodes in seven patients, showing trajectories traversing both the NAc and ALIC. ALIC, anterior limb of internal capsule; NAc, nucleus accumbens. Local gamma oscillations in the NAc-ALIC may as a key biomarker linked to depressive symptoms Gamma oscillations (30–90 Hz) are implicated in various cognitive processes, including learning,33 memory,34 and attention,35 and are closely associated with multiple psychiatric disorders.36,37,38 They are regarded as a significant biomarker of depression, correlating with both symptom severity and treatment responsiveness.39,40 Notably, responders to conventional antidepressants or ketamine frequently exhibit enhanced gamma activity. Previous studies have also reported that gamma oscillations in the right lateral habenula are negatively correlated with preoperative scores on the HAMD and HAMA scores.41 The fundamental alterations in gamma oscillations are thought to originate from an imbalance between inhibitory and excitatory neuronal activity in the brain.42,43 As proposed by Marin et al., the maintenance of normal neural circuit function and stability depends on a “transient balance” between excitatory inputs from glutamatergic neurons and inhibitory inputs from GABAergic neurons.44 The generation of gamma oscillations in the NAc is influenced by synaptic inputs from several brain regions, such as the prefrontal cortex, amygdala, hippocampus, ventral tegmental area, and thalamus.45 The NAc functions as an “information switch”, not merely receiving inputs passively but actively selecting and integrating dominant information for output.46 Under stress, high gamma oscillations in the NAc can synchronize with other regions exhibiting similar oscillatory activity, such as the prefrontal cortex, thereby influencing behavioral outcomes. Stress and distress are associated with changes in synaptic transmission and structural volume in cortical and limbic areas, including the prefrontal cortex and hippocampus.47,48 These alterations lead to a decline in gamma oscillations and disrupt information exchange between the NAc and other brain regions.49 DBS-mediated theta network synchronization as a system-level mechanism of symptom improvement Beyond local modulation, we observed that DBS enhanced theta oscillations across distributed cortical and limbic regions, including the left orbitofrontal cortex, bilateral medial frontal cortex, left limbic system (entorhinal cortex, parahippocampus, and temporal pole), left parietal cortex, and right occipital cortex. Theta activity is known to be generated in prefrontal-limbic-striatal circuits50,51 and is closely linked to memory and emotional regulation.52,53 Its increase after DBS has been previously associated with clinical improvement in SCC-DBS studies.54 At present, it was believed that the function of the limbic system-cortex-striatum-pallidum-thalamus circuit was abnormal in depressed patients.55 The cortex mainly involved the anterior cingulate cortex, orbitofrontal cortex, and SCC.56 Lynch et al. found that the frontal-striatal network in patients with depression expanded significantly, especially in the frontal cortex.57 The frontal lobe and the striatum were closely related in the pathogenesis of depression. Based on current neuroscientific models, NAc-ALIC-DBS likely drives fronto-limbic theta synchronization through a multi-stage mechanism. First, high-frequency stimulation resets pathological network activity via axonal activation and synaptic inhibition,58 while simultaneously inducing local gamma oscillations that are harmonics of the stimulation frequency.59,60 This gamma activity may engage in cross-frequency coupling with lower frequency rhythms, modulating the response gain of ventral striatal neurons to theta-rhythmic inputs from upstream regions like the prefrontal cortex and hippocampus, thereby enhancing the theta rhythmicity of striatal output.61,62 This theta rhythm then propagates through the “limbic-cortex-striatum-pallidum-thalamus” circuit, leveraging the intrinsic oscillatory properties of this loop to entrain the entire network toward a common low-frequency resonant point.63 Ultimately, this process enhances theta-band phase synchrony between the prefrontal cortex and limbic structures such as the amygdala and hippocampus, optimizing information integration within emotion-cognition circuits.64,65 An early marker of this process may be changes in oscillatory power within specific frequency bands, such as theta. The initial therapeutic mechanism of DBS in this disorder likely involves modulating the excitation-inhibition balance and oscillatory stability within local neural circuits, rather than directly reshaping directed information flow across large-scale brain networks, that is also the reasons for the lack of positive results in granger causality analysis. The increase in theta activity may reflect enhanced local inhibitory control or network stabilization, which alone can lead to clinical symptom relief.66 Programmable stimulation Stimulation parameter setting is crucial for DBS therapy. The adjustment process is time-consuming and requires close doctor-patient communication, as well as patience and persistence. Multiple rounds of adjustment are often needed to identify the optimal parameters, which also explains why large-scale research in this area remains challenging. Electrode contact selection Acute stimulation of the NAc immediately alleviated anhedonia in patients with TRD, whereas ALIC stimulation showed no such effect. The duration of pleasure maintenance varied across individuals. In one patient, the effect diminished after one week; in three patients, it lasted for three months; and in the remaining three patients, it persisted for one month. Subsequent enhancement of stimulation by adding extra contacts may improve outcomes. One feasible strategy is to stimulate one contact within the NAc while simultaneously targeting another in the ALIC. The optimal combination of these contacts should be adjusted based on individual patient response, with specific parameters detailed in the supplementary materials (Table S1). Among the seven patients who received NAc stimulation, five were additionally stimulated at the ALIC. Notably, two patients met the response criteria with NAc stimulation alone. The observed outcomes may be attributed to the following factors: first, stimulation of the NAc likely constitutes the predominant therapeutic mechanism, as its induced electric field often extends to adjacent nuclei, including the BNST, which is associated with enhanced antidepressant effects.67,68 Second, the two patients exhibited a relatively short illness duration and a symptom profile primarily characterized by anhedonia, and NAc stimulation alone was sufficient to modulate dysfunctional emotional circuitry. In contrast, patients with more severe symptoms and longer illness durations may require combined NAc and ALIC stimulation to more comprehensively restore abnormal neural pathways and optimize treatment efficacy. However, it should be noted that patients with prolonged illness durations (exceeding ten years) often show a diminished response to DBS, suggesting that long-standing dysfunction in emotion-related circuits may be less amenable to restoration through neuromodulation. Voltage, pulse width, and frequency Long pulse width stimulation showed a wider range and more significant therapeutic effects.69,70 Patients who did not respond to short pulse width treatment showed efficacy with long pulse width, possibly due to greater current diffusion affecting more abnormal neural circuits.71 In long-term stimulation, 130 Hz stimulation was more effective than 20 Hz stimulation, with no significant difference in side effects between the two frequencies.72 The stimulation frequency used in this study was mostly 130 Hz, similar to the stimulation frequency reported in existing literature.73 Based on our experience, the initial stimulation voltage should be set cautiously. A range of 2.5–4 V is generally considered safe. In cases where patients did not respond adequately to lower voltages, a gradual increase may enhance therapeutic outcomes. Nevertheless, close monitoring for adverse events is essential throughout voltage adjustment. Some studies have also reported that no side effects were observed even under 9 V or 10 V.74 Limitations and future directions This study had a small sample size (n = 7) and lacked the control group, a common limitation in research on DBS for treatment-resistant psychiatric disorders. Limited patient awareness of DBS as a treatment option and its high cost contributed to recruitment challenges. Moreover, long-term clinical effectiveness needs further validation. This study provided a snapshot of network modulation and did not include longitudinal MEG and LFP data. Therefore, we could not draw conclusions about the persistence of the observed “theta-specific network” effects beyond the study period. It is necessary for longitudinal MEG/EEG studies that track theta-network dynamics in DBS patients over years, correlating changes with clinical state. Additionally, this study only focused on resting-state LFPs and MEG, future work will explore electrophysiological changes across task states (e.g., emotional processing) and sleep states, integrating LFPs, EEG, and MEG data for a more comprehensive biomarker analysis. Self-control methods, such as DBS withdrawal, posed risks for TRD patients, many of whom experienced severe distress and suicidal behavior. Sham stimulation may be more feasible once parameter optimization is completed and the patient’s condition has stabilized. Future studies could enhance the translational potential of neuromodulation strategies by systematically recording LFPs, MEG, or EEG before and after altering stimulation parameters—particularly when switching between targets such as the NAc and the combined NAc+ALIC. This approach would allow for a more rigorous analysis of stimulation-induced neurophysiological changes, potentially leading to the identification of robust biomarkers for optimizing therapeutic outcomes. Larger sample sizes and control groups (such as including a conventional treatment group) will be valuable for validating our findings. Resource availability Lead contact Further information and requests for resources should be directed to and will be fulfilled by the lead contact Wei Wang (wcnsww@163.com). Materials availability The study did not generate new unique reagents. Data and code availability •Data were obtained from publicly available datasets by Zhang et al. (https://www.doi.org/10.57760/sciencedb.34841).•This paper does not report original code.•Any additional information required to reanalyse the data reported in this paper are available from the lead contact upon request. •Data were obtained from publicly available datasets by Zhang et al. (https://www.doi.org/10.57760/sciencedb.34841). Data were obtained from publicly available datasets by Zhang et al. (https://www.doi.org/10.57760/sciencedb.34841). •This paper does not report original code. This paper does not report original code. •Any additional information required to reanalyse the data reported in this paper are available from the lead contact upon request. Any additional information required to reanalyse the data reported in this paper are available from the lead contact upon request. Acknowledgments This work has been supported by grants from Clinical Research Incubation Project, 10.13039/501100013365West China Hospital, Sichuan University (no. 2022HXFH024); Zero to One Project, Sichuan University (no. 2022SCUH0035); Achievement Transformation Project, West China Hospital of Sichuan University (no. HX20190281); STI 2030—Major Projects(no. 2021ZD0200407). All electrodes in this study were supported by 10.13039/501100015053SceneRay. Author contributions Conceptualization, data curation, methodology, software, investigation, validation, visualization, formal analysis, writing – original draft, writing – review and editing, W.Z.; conceptualization, data curation, methodology, software, visualization, investigation, writing – original draft, writing – review and editing, B.X.; conceptualization, data curation, methodology, investigation, resources, visualization, writing – original draft, Y.K.; conceptualization, investigation, project administration, Y.G.; software, data curation, validation, R.Y.; conceptualization, resources, project administration, supervision, Q.W.; conceptualization, funding acquisition, resources, supervision, writing – review and editing, W.W. All authors have read and agreed to the published version of the manuscript. Declaration of interests The authors declare that they have no competing interests. STAR★Methods Key resources table REAGENT or RESOURCESOURCEIDENTIFIERDeposited dataRaw dataThis paperhttps://www.doi.org/10.57760/sciencedb.34841Software and algorithmsPrism 10GraphPadhttps://www.graphpad.com/Adobe Illustrator CC2020Adobehttps://www.adobe.com/cn/SPSS 22.0IBMhttps://www.ibm.com/cn-zh/products/spss-statisticsPython (v3.8)PSFhttps://www.python.org/SPM12UCLhttps://www.fil.ion.ucl.ac.uk/spm/ REAGENT or RESOURCESOURCEIDENTIFIERDeposited dataRaw dataThis paperhttps://www.doi.org/10.57760/sciencedb.34841Software and algorithmsPrism 10GraphPadhttps://www.graphpad.com/Adobe Illustrator CC2020Adobehttps://www.adobe.com/cn/SPSS 22.0IBMhttps://www.ibm.com/cn-zh/products/spss-statisticsPython (v3.8)PSFhttps://www.python.org/SPM12UCLhttps://www.fil.ion.ucl.ac.uk/spm/ Experimental model and study participant details Human participants This prospective, open-label, single-center clinical trial was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Review Committee of West China Hospital (No.2022-419). The work has been reported in line with the STROCSS criteria.75 Patients who met the inclusion and exclusion criteria were enrolled. Written informed consent was obtained from all participants and their legal guardians. Given that adolescence represents the peak time for the onset of depression, the brain structure and function of individuals aged 16 be prone to reach maturity and stability,76 the minimum age threshold for patients participating in this study was set as 16 years old.77,78 Inclusion criteria: Age between 16 and 65 years, Male or female; Diagnosis of depressive disorder according to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-V); HAMD-24 score ≥20; Duration of depressive episode ≥ two years or recurrent episodes, with the current episode lasting for at least one year; Failed to respond to at least three adequate trials of antidepressants with different mechanisms, provided in adequate doses and duration, each administered at an adequate dose for a minimum treatment duration of 8 weeks; Refused or poor response to electroconvulsive therapy; The participant possessed general cognitive and communication abilities; Both the participants and their legal guardian were obligated to sign a written, informed consent form. Exclusion criteria: Diagnosis of bipolar disorder, schizophrenia, severe personality disorders or other psychiatric disorders; Serious physical illness or neurological disorders; Alcohol or substance use disorder; Preoperative cranial magnetic resonance imaging (MRI) reveals structural abnormalities in the brain; Contraindications for MRI examination; Pregnancy or lactation; Participants were enrolled in another research; After a comprehensive evaluation by the research team, the patient was ineligible for participation in this clinical trial. Method details Surgical plan and clinical assessment The 4-contact DBS electrodes (Model 1242, SceneRay) were implanted along a trajectory designed to traverse the ALIC and terminate within the NAc (Figure 3A). The three-dimensional renderings are presented in Figure 3B, and the production details was in previous article.67 Details of the target were provided in supplementary materials (Figure S3). Two experienced psychiatrists who were blinded to the treatment conditions conducted a comprehensive evaluation of TRD patients, gathering essential demographic information and medical history. Psychiatric scales assessed in the study included the HAMA, HAMD-24, DARS, TEPS, MCCB, WHOQOL, GAF, and PSQI. The primary outcome was the severity of the depressive symptoms assessed by HAMD-24. The reduction of HAMD-24 score by at least 50% was defined as response, whereas remission was defined as < 8 points on the HAMD-24 score. The reduction rate ranging from 25% to 50% was defined as partial response. Other scales were used as secondary outcomes. Patients met with the study psychiatrists preoperatively, one month, three months, six months after DBS and received clinical assessment. Six months after initiation of DBS was defined as follow-up endpoint. Follow-up would persist according to the patient’s cooperation. Stimulation adjustments were made based on the patient’s conditions. The programming process was detailed in the supplementary materials. LFPs recording and processing LFPs from the NAc and ALIC were recorded monthly using the SceneRay software (SceneRay Co., Ltd., China) with stimulation temporarily turned off. Recordings were obtained while participants remained seated at rest with eyes open, refraining from movement and speech. Immediately following each LFP recording session, the HAMD-24 was administered to assess depressive symptom severity. Bipolar recordings were obtained from the following contact pairs: left NAc (contacts 0–1), left NAc–ALIC junction (contacts 1–2), left ALIC (contacts 2–3), right NAc (contacts 4–5), right NAc–ALIC junction (contacts 5–6), and right ALIC (contacts 6–7). Data were acquired at a sampling rate of 1000 Hz, with each session lasting approximately 3 min per contact pair. Raw signals were first processed through a 2nd-order Butterworth band-pass filter (1–100 Hz) to retain physiologically relevant frequency bands. Subsequently, to eliminate power line interference, 2nd-order Butterworth notch filters were applied to attenuate 40 Hz and 80 Hz line noise artifacts. All filtering operations were implemented using zero-phase filtering to avoid phase distortion. Power spectral density (PSD) was estimated using Welch’s method with a 1-s Hanning window, 50% overlap, and 2048 FFT points. Frequencies below 1 Hz were excluded from analysis due to high-amplitude drift artifacts. The PSD was decomposed into six canonical frequency bands: delta (1–4 Hz), theta (5–8 Hz), alpha (9–12 Hz), beta (13–30 Hz), low-gamma (31–60 Hz), and high-gamma (61–90 Hz) (Figure 2A). MEG recording and processing Resting-state MEG data were acquired using a 306-channel Elekta Neuromag TRIUX system (Elekta Oy, Finland) prior to DBS implantation and one month, three months, and six months post-operatively. During recording, participants lay supine in a magnetically shielded room with eyes closed. Continuous data were collected for 10 min at a sampling rate of 2000 Hz. Preprocessing included visual inspection and removal of artifacts related to head movement, eye blinks, and muscle activity, channels with excessive noise were excluded, data were then band-pass filtered (1–100 Hz) and notch-filtered at 50 Hz . The source localization encompassed three core stages: head model construction, inverse source estimation, and spatial normalization. First, individual high-resolution T1-weighted MRI were segmented into scalp, skull, cerebrospinal fluid, and gray/white matter compartments using SPM12 to generate tissue surfaces. Scalp and electrode positions from a standard cap layout were then co-registered to the individual MRI using fiducial landmarks and an iterative closest point algorithm. A three-layer Boundary Element Method (BEM) head model was constructed from these surfaces to compute the lead field matrix. Second, the inverse problem was solved using the dynamic Statistical Parametric Mapping (dSPM) approach. Cortical current density was estimated via a linear weighted minimum-norm estimate (wMNE) on a dense cortical source space (∼10,000 vertices), with regularization parameter (λ) determined per subject using the L-curve method and adjusted based on signal-to-noise ratio. Finally, for group-level analysis, individual source estimates were morphed to the FreeSurfer fsaverage template, yielding time-varying neuronal current amplitude at each cortical vertex. Cortical source activity was estimated using the Desikan-Killiany atlas, PSD and Granger-causality connectivity were computed for the same six frequency bands described above (Figure 2B). The aforementioned steps were implemented using modified and custom Python (v3.8) scripts integrated with the Brainstorm toolbox. Quantification and statistical analysis Normality of paired samples was assessed using the Shapiro–Wilk test. For non-normally distributed data, the Wilcoxon signed-rank test was applied; for normally distributed data, the paired t test was used. Multiple-comparison correction was performed using the false discovery rate (FDR) method. Associations between HAMD-24 scores and LFP spectral features were evaluated via spearman’s correlation test. Statistical significance was defined as p < 0.05 (two-tailed). Additional resources The study has been registered in Clinical Trial Registry (https://www.chictr.org.cn/) with the ID: ChiCTR2300069983.