Heart Rate Variability Biofeedback May Reduce Mild Depression by Changing Brain-Heart Communication Through the Habenula Network

A total of 34 university students with StD and 32 HC were recruited from Guangzhou Xinhua University. The two groups were matched for age and education level. Among the StD participants, 16 (15 females, 1 male) voluntarily underwent HRV‐BF training, of whom 9 (8 females, 1 male) completed resting‐state fMRI scans before and after the intervention. Additionally, 17 HC subjects completed one fMRI scan (Figure 1).

This study was conducted according to the Declaration of Helsinki. Ethical approval was obtained from the biological and medical ethics committee of Guangzhou Xinhua University. All participants gave informed consent.

Demographic data, including age, sex, and body mass index (BMI), were collected using standardized questionnaires. Depression severity was assessed by two psychiatrists or trained psychological counselors using the 17‐item Hamilton Depression Rating Scale (HAMD‐17) [21] and the Patient Health Questionnaire‐9 (PHQ‐9) [22]. Sleep quality was evaluated with the Pittsburgh Sleep Quality Index (PSQI) [23]. The scoring criteria for the scales are provided in the Supporting Information—Method Table S1.

Pulse waves were recorded from the non‐dominant index finger using a photoplethysmographic (PPG) sensor. R–R intervals were extracted and analyzed for HRV using the BioTrace software integrated with the NeXus‐32 biofeedback system (Mind Media, the Netherlands).

Blood pressure was measured in the brachial artery of the right upper arm using an automatic electronic sphygmomanometer (Omron HEM‐8102 K). Two consecutive readings were taken 1–2 min apart, and their average was used as the final blood pressure value. If the difference between the two systolic blood pressure (SBP) readings exceeded 10 mmHg or the difference in diastolic blood pressure (DBP) exceeded 5 mmHg, a third measurement was taken, and the average of the two closest readings was used.

The training protocol consisted of a pre‐training phase followed by formal training (details are provided in the Supporting Information—Method). Training was conducted in a quiet room with ambient temperature maintained between 24°C and 26°C. All participants were instructed to avoid consuming stimulants such as strong tea, coffee, or alcoholic beverages within 24 h prior to training. 1.2

All participants underwent MRI scanning using a 3.0 T Siemens Prisma scanner (detailed parameters in Supporting Information—Method). During scanning, subjects wore noise‐reducing headphones, were instructed to avoid active thinking, and remained awake with eyes closed while lying still on the scanner bed. 1.3.1

Statistical analyses were performed using SPSS version 29.0. Continuous variables conforming to normal distribution were expressed as mean ± standard deviation (mean ± SD) and compared using independent samples t‐tests. Non‐normally distributed variables were described as median (interquartile range) [M(P₂₅, P₇₅)] and compared with the Mann–Whitney U test. Paired samples were analyzed using paired t‐tests if normally distributed, or Wilcoxon signed‐rank tests if not. Categorical variables were compared with Pearson's chi‐square (χ²) tests. All tests were two‐tailed, with p < 0.05 considered statistically significant. Resting‐state fMRI preprocessing and functional connectivity analyses were conducted using RESTplus (v1.30) (detailed in Supporting Information—Method 1.3.6).

There were no significant differences between the StD and HC groups in age, sex, or BMI (Table 1). However, scores on clinic scales showed that StD participants had significantly higher scores on the HAMD‐17, PHQ‐9, and PSQI compared to HCs (p < 0.05).

Compared to the HC group, participants in the StD group exhibited elevated HR (Figure 2c; p < 0.001), while no significant differences were observed in SBP or DBP (Figure 2a,b; p > 0.05).

HRV analysis revealed that the StD group had significantly reduced lnRMSSD, HF power (Figure 2e,f; p < 0.01), along with elevated LF power (Figure 2g; p < 0.001). The lnSDNN was significantly decreased (Figure 2d; p < 0.05), and HRV amplitude showed a decreasing trend (Figure 2h; p = 0.055), reflecting a potential reduction in overall autonomic regulation. These findings suggest that individuals with StD exhibit autonomic imbalance characterized by decreased vagal tone and relatively increased sympathetic activity compared to HCs (see Table S2) for full statistics.

Voxel‐wise whole‐brain functional connectivity analysis using bilateral Hb as seed regions revealed that, compared to the HC group, the StD group exhibited significantly reduced functional connectivity between bilateral Hb and multiple brain regions, including precuneus and right posterior cingulate cortex (Default Mode Network, DMN); left middle cingulate cortex (Salience Network, SN); right superior parietal lobule(Central Executive Network, CEN); cuneus, right calcarine cortex, and left superior occipital gyrus(Visual Network, VN); right paracentral lobule and right postcentral gyrus(Somatomotor Network, SMN) (Figure 4, left panel; Table 2). These findings suggest that in individuals with StD, the habenula exhibits reduced functional connectivity with these regions may contribute to alterations in self‐referential processing, sensorimotor integration, and visuospatial attention.

After 4 weeks of HRV‐BF, participants with StD exhibited significant improvements in vagal‐mediated HRV parameters. Specifically, lnRMSSD and HF power significantly increased (Figure 5e,f; p < 0.05), along with lnSDNN (Figure 5d; p < 0.05). In contrast, LF power significantly decreased (Figure 5g; p < 0.05). No significant changes were observed in HRV amplitude (Figure 5h), SBP, DBP, HR (Figure 5a–c), or respiratory rate (Figure 5i). These results suggest that HRV‐BF increases vagal tone and restores autonomic balance.

Post‐intervention results showed significant reductions in HAMD‐17, PHQ‐9, and PSQI scores (Figure 6a–c; p < 0.05), indicating substantial improvement in depressive symptoms and sleep quality. Neuroimaging data demonstrated significantly enhanced positive functional connectivity between the bilateral Hb and several reward‐affective‐motor‐related ROIs, including the Hb‐NAc (Figure 6d; p < 0.01), Hb‐VeP (Figure 6e; p < 0.01), Hb‐AMY (Figure 6f; p < 0.01), and Hb‐GPi (Figure 6h; p < 0.05). Among these, the upregulation of Hb‐NAc functional connectivity was the most robust (Figure 6i). Hb‐SN functional connectivity (Figure 6g) and Hb‐other ROIs showed no significant changes (see Figure S2). Furthermore, voxel‐wise analysis of whole‐brain Hb‐FC revealed post‐intervention increases in functional connectivity with regions including: precuneus (DMN); left middle cingulate cortex (Salience network); bilateral paracentral lobule and left supplementary motor area (SMN) (Figure 4), right panel; (Table 2). These findings suggest that HRV‐BF may alleviate depressive symptoms by modulating dysfunctional Hb‐centered brain network functional connectivity.

The brain–heart axis is a bidirectional regulatory network comprising the CAN, the intrinsic cardiac nervous system (ICNS), and the ANS [10]. CAN is the core component of the brain–heart axis, integrating information from cortical and subcortical regions, and dynamically regulating sympathetic and parasympathetic outflow via brainstem cardiovascular centers such as the rostral ventrolateral medulla (RVLM), nucleus tractus solitarius (NTS), dorsal motor nucleus of the vagus (DMV), and nucleus ambiguus (NA), etc. Through these descending pathways, the CAN modulates the ICNS to maintain cardiac autonomic homeostasis [10]. Chronic stress or depressive states can disrupt this regulation, resulting in cardiac autonomic dysregulation [28].

In this study, StD participants exhibited decreased vagal tone and increased sympathetic activity (Figure 2), indicating impaired descending regulation of the brain–heart axis. To investigate underlying central mechanisms, we examined Hb‐related functional connectivity using BOLD rs‐fMRI. Results showed aberrant Hb‐centered network connectivity in StD, specifically involving the NAc, VeP, AMY, SN and GPi (Figure 3). Voxel‐wise analysis revealed reduced functional connectivity between the Hb and the precuneus, right posterior cingulate cortex, left middle cingulate cortex, right superior parietal lobule, cuneus, right calcarine cortex, left superior occipital gyrus, right paracentral lobule, and right postcentral gyrus (Figure 4 and Table 2). Many of these regions are key nodes of large‐scale functional networks—including the default mode network (DMN), salience network (SN), central executive network (CEN), sensorimotor network (SMN), and visual network (VN). Reduced functional connectivity between Hb and these nodes may contribute to alterations in self‐referential processing, sensorimotor integration, and visuospatial attention, partially consistent with previous reports [29, 30]. Collectively, these findings indicate that disrupted Hb‐centered functional connectivity may contribute to brain–heart axis dysfunction in StD.

The LHb serves as a central hub for processing negative emotions and inhibiting rewards. To further elucidate how Hb‐centered network abnormalities influence cardiac autonomic regulation, we selected 13 brain regions anatomically connected to the LHb as regions of interest (ROIs) and examined their functional connectivity patterns [27]. Our results demonstrated that with increasing depressive severity, the functional connectivity between the Hb and several reward‐, affective‐, and motor‐related regions (NAc, VeP, AMY, SN, GPi) shifted from coupling to anti‐coupling (Figure 3). These regions maintain close structural and functional associations with the CAN [31, 32, 33, 34]. Functional anti‐coupling within these circuits may disrupt the integration between reward‐, affective‐, motor‐, and autonomic systems, thereby weakening adaptive regulation of the brain–heart axis.

Importantly, these five regions are also tightly linked to the midbrain monoaminergic system, including the ventral tegmental area (VTA) and the dorsal raphe nucleus (DRN) [34, 35, 36]. As an upstream regulator, the LHb inhibits dopaminergic (DA) neurons in the VTA and serotonergic (5‐HT) neurons in the DRN, thereby reducing monoaminergic output [27]. We therefore hypothesize that by suppressing these monoaminergic nuclei, the LHb reduces excitability in reward‐related, affective‐related, and motor‐related regions. Consequently, this cascade of events weakens functional integration with the CAN and ultimately impairs the central regulation of cardiac autonomic function.

Although this hypothesis was not directly tested in the present study, prior animal experiments provide strong supporting evidence: Electrical stimulation of the LHb increases mean arterial pressure (MAP) and decreases heart rate (HR), effects that are attenuated by DA and 5‐HT receptor antagonists or VTA lesions, suggesting that the LHb can modulate cardiovascular activity via the monoaminergic system [37, 38].

Notably, we did not directly observe functional connectivity abnormalities between Hb–VTA or Hb–DRN in the current study. However, previous research indicates that Hb–VTA functional connectivity anti‐coupling emerges as depression progresses to MDD [39], suggesting that functional disruption in the StD stage may still be at an early or compensatory stage, not yet manifesting as pronounced functional connectivity anti‐coupling. Moreover, ROC analysis showed that Hb–NAc and Hb–VeP functional connectivity effectively discriminated StD from HC (AUC = 0.876, p < 0.001), highlighting their potential as neuroimaging biomarkers with clinical relevance.

Collectively, these findings suggest that Hb‐centered network abnormalities in StD may impair integration between reward‐, affective‐, and motor‐regulating circuits and autonomic control centers, resulting in dysregulated brain–heart axis activity and cardiac autonomic imbalance.

Our previous findings indicated that central regulation of the brain–heart axis is impaired in StD, manifested as Hb‐centered network abnormalities and reduced vagal tone. Given the bidirectional nature of the brain–heart axis, peripheral autonomic changes can influence central network plasticity and function through automatic afferent pathways, in addition to the descending control from the CNS [10]. Therefore, we hypothesized that enhancing vagal tone peripherally might restore impaired Hb‐centered network function and rebalance the brain–heart axis.

Heart rate variability biofeedback (HRV‐BF) provides a feasible approach to this hypothesis. As a non‐pharmacological neuromodulation technique, HRV‐BF enhances vagal activity through rhythmic breathing training at a resonance frequency (~0.1 Hz). Mechanistically, HRV‐BF increases vagal tone, enhancing signal transmission from the NTS to the CAN [19], thereby regulating sympathetic–parasympathetic balance. However, direct evidence demonstrating that HRV‐BF modulates Hb networks to remodel brain–heart axis integration has been lacking.

In the present study, after 4 weeks of HRV‐BF intervention, StD participants exhibited significantly increased vagal tone, with cardiac autonomic function shifting from sympathetic dominance toward parasympathetic predominance, indicating restored autonomic function (Figure 5). Concurrently, functional connectivity between the Hb and reward‐, affective‐, and motor‐related regions (NAc, VeP, AMY, GPi) was strengthened (Figure 6). Additionally, enhanced coupling was observed between the Hb and key nodes of large‐scale networks, including the DMN, Salience network, and SMN (Figure 4, right panel; Table 2). These results suggest that HRV‐BF may enhance Hb network function via “bottom‐up” vagal modulation. Restoration of Hb connectivity may in turn promote “top‐down” regulation of reward‐, affective‐, and motor‐related circuits with autonomic centers, forming a closed‐loop mechanism for brain–heart axis integration.

This study has several limitations. First, the sample size was relatively small and female‐dominant (93.75%). While this reflects the higher prevalence of StD in women, it may limit the generalizability of the findings to male populations. Second, the 3.0 T MRI resolution precluded precise separation of LHb and MHb subregions, though selecting 13 ROIs anatomically connected to the LHb partially mitigated this limitation. Third, long‐term effects of HRV‐BF were not assessed, as the study focused on immediate changes after 4 weeks, though prior studies suggest sustained benefits [18, 40, 41]. Fourth, inter‐rater reliability was not formally calculated, but standardized training likely minimized scoring bias. Fifth, no active control group was included. A pre–post design in our study captured individual changes, and although placebo effects cannot be ruled out, the findings provide preliminary evidence supporting HRV‐BF's potential benefits. Future studies should recruit larger, gender‐balanced cohorts, incorporate active control interventions, include long‐term follow‐up, and utilize multimodal neuroimaging techniques (e.g., DTI, PET) to further elucidate the neural mechanisms and support clinical translation of HRV‐BF in subthreshold and major depression.