Integrated biofeedback (BF) and virtual reality (VR) technologies have been widely applied in recent years [1,2]. Accumulating evidence has shown that the application of integrated BF and VR technologies may mitigate users’ stress levels [3,4], anxiety [4-6], self-reported pain [7,8], and developmental disorders [9,10], and improve people’s cognitive function [11,12]. Previous interventions for psychological disorders have employed a variety of BF modules, including electroencephalography (EEG), electromyography, skin conductance, heart rate, heart rate variability, body temperature, and respiration [13-19]. Integrated VR and BF technologies are especially common for daily self-help training and breathing training [4,20]. For example, VR and BF have been used to promote diaphragmatic breathing [3] and stress regulation with breathing training among police officers [12]. Breathing training with VR and BF technologies shows a promising future in terms of cultivating the mental health of the population, and the understanding of the mechanism underneath breathing training is essential to popularize breathing training among the public [1,3,12].

VR and BF address the challenge of daily training engagement adequately [3,12]. The development of VR is blooming because its interactive nature and gamification elements facilitate self-help training and home-based rehabilitation [21,22]. Self-help training encourages people to engage in self-training in their own time to mitigate illness. Still, a high dropout rate often challenges self-help training because the training is often perceived as boring with no immediate excitement for the participants [4,20,23]. As a new-generation human-computer interface, VR provides an immersive user experience with its 360° visual environment [12,24], thus offering excitement and motivating users to continue training [6,25]. Moreover, VR applications can integrate gaming elements into training, such as storylines, roleplay, progression systems, and rewards [21,22]. VR-BF integration also addresses users’ concerns that training content is boring and meaningless, and thus facilitates training consistency [4,20,23]. Overall, integrated VR and BF interventions provide users with interactive experiences, involvement, and excitement [4,20,23]. For instance, intervention sessions combining VR with BF were developed to improve body coordination and balance in young adults with juvenile idiopathic arthritis [26] and children with cerebral palsy [27], and to improve cognitive function under stressful circumstances in police offices [12]. The fundamental reason behind the effectiveness of VR-BF integrated training is that these technologies allow more effective self-help training by supporting trainees’ cognitive engagement [2,11]. Studies suggested that VR increased people’s engagement with self-help training among both adults and children, as participants reported pleasurable training experiences [11,28,29]. On the other hand, BF (eg, breathe monitoring) increased trainees’ awareness toward their breath, therefore allowing them to regulate their breath if needed [3,23]. BF can further reinforce trainees’ engagement through the use of feedback stimuli, and the gamification design further increases the attractiveness of breathing training [4,23,30,31], as game-like training offers the possibility to facilitate trainees’ immersion and boost engagement [30,32]. Overall, evidence supports that the combined use of VR and BF is a way to increase the engagement of self-help training among trainees.

Clinically, a growing population of therapists has combined VR-BF technologies with breathing training [1,3,33]. Breathing training is a common intervention in self-help training, health care, and mindfulness practice [34,35]. Breathing training interventions have been shown to mitigate stress, pain, anxiety, and depression, and improve relaxation [36-40]. Research has also shown that breathing training can be integrated with breathing signal BF and VR to promote users’ heart function and heart rate variability [1,41]. At the consumer level, many smartphone or smartwatch applications provide different breathing training programs with heart rate or breath rate BF [42,43]. In fact, the quantitative and qualitative feedback from VR and BF breath intervention trainees suggested that VR and BF technologies provide positive, relaxing, and peaceful user experiences [31,44]. Underneath the behavioral benefits of breathing training, review studies suggested that breathing training modulates the vagal tone of the parasympathetic nervous system and thereby increases heart rate variability and respiratory sinus arrhythmia in trainees [39,45,46]. These physiological changes further initiate an autonomic relaxation response and regulate the autonomic nervous system to achieve a balance state [3,47]. Overall, both behavioral and physiological evidence suggests that integrating breathing training with BF and VR is a promising strategy to promote personal health in trainees.

Despite behavioral and physiological evidence, some scientists attempted to study what is going on inside of breath trainees’ brains from the neurological perspective [39,48]. EEG studies on breathing training showed an increase in alpha band power in the prefrontal area during Zen Tanden breathing [49,50] and in the occipital and right temporal areas during mindful breathing [51]. A study also found decreased theta band power in the left frontal, right temporal, and left parietal areas when people practice paced breathing training [52], but other studies reported increased theta band power in the temporal, parietal, central, and occipital areas in mindfulness practitioners and in alternate nostril yoga breathing practitioners [48,51]. Studies also reported increased beta and gamma power in the frontal area and the midline in mindful breathing practitioners, yoga breathing practitioners, and Rinpoche, a group of highly experienced Tibetan Buddhism meditators [53-56], while another study on alternate nostril yoga breathing practitioners showed decreased beta band power [48]. The gamma band effective connectivity also increased in people who received 8 weeks of mindful breathing training [57]. Currently, studies on various breathing training techniques coexist and the neurological evidence of them is inconsistent [11,39,58]. The neurological elaboration of different breathing training techniques is essential to the understanding of the nature of breathing training. Yet, this neurological knowledge gap is unaddressed.

We reviewed previous studies on breathing training and identified 3 basic breath techniques that were mostly employed: mindful breathing, guided breathing, and breath counting [33,59,60]. For example, some mindfulness-based interventions promoted changes in trainees through mindful breathing, emphasizing nonjudgmental awareness of people’s body sensations [59,61]. Another intervention, breath counting, instructed participants to count from 1 to 10, along with their breath [60,62]. Lastly, guided breathing interventions instructed participants to breathe at a certain and slow tempo [33,63]. Although studies supported that these breathing techniques would introduce certain benefits to trainees [33,59-63], the differences in their neurological mechanisms are still largely unknown. We attempted to address this neurological knowledge gap in this study. The understanding of the neurobiological mechanisms underlying different breathing techniques is important for the development of VR-BF breathing training and may unveil our knowledge to human brain neuroscience.

In this study, we attempted to use EEG to examine the neurobiological mechanism differences of different breathing training techniques in an integrated VR and BF environment. EEG is commonly applied in breathing training studies, as it can easily be integrated with other modalities, such as VR and breath signal sensors [13]. Previous EEG studies on breathing training with VR and BF also demonstrated that EEG suitably works with VR-BF equipment without significant interference [10,64]. The research questions are whether and how people’s neurological mechanisms (indicated by EEG indexes) change when they are performing different breathing training techniques (ie, mindful breathing, guided breathing, and breath counting). This study aimed to examine the EEG indexes systematically, including band power, effective connectivity inflow and outflow, and connectivity flow among different brain regions, and analyze the differences of these indexes when people are practicing different breathing training techniques. This comparison will not only reveal the neurological mechanisms of these techniques, but also verify the feasibility of applying integrated VR and BF techniques for breathing training in healthy populations, like the consumer-level digital products proclaimed (eg, smartphones) [41,65].

The data were collected between the fall of 2021 and the spring of 2022. The researchers recruited potential participants through online forum advertisements and billboard posters. The inclusion criteria were as follows: age ≥20 years; no previous mindfulness training experience; no history of epilepsy, mental disorder, or neurological illness; and no history of substance abuse (alcohol, drugs, or tobacco). Fifty-three candidates from 2 universities in northern Taiwan who fulfilled the above criteria were included in this study. The data of 2 participants were excluded because of extreme data deviation. Eventually, the analyses included the data of 51 young participants (26 men and 25 women; mean age 22.7 years, SD 2.5 years; range 20-29 years).

National Yang Ming Chiao Tung University – Research Ethics Center for Human Subject Protection (NYCU-REC) reviewed and approved this study (project number: NCTU-REC-109-037F). All participants provided written informed consent before participation. The participants received New Taiwan Dollar 1000 (about US $32.69) in acknowledgement of their participation. In accordance with the regulations of NYCU-REC, all the research materials and data were secured in the designated location, which was protected from any unauthorized personnel by 3 locks. Only project-related personnel, as listed in NYCU-REC (project number: NCTU-REC-109-037F), could access the data. Only the experiment manipulators and the contact persons between NYCU-REC and the laboratory could access the participants’ personal information. After the experiments were finished, every participant’s identity was replaced by a subject number and deidentified. No material that may identify any of the participants is presented in this paper.

The experiment manipulators explained the experimental procedure to the participants and equipped them with an EEG system, an electrocardiography (ECG) sensor, a respiration belt, and a VR headset. The top-left corner of Figure 1 shows the configuration of equipment on a participant during the experiment.

An HTC Vive VR headset was used to create the VR environment, as shown in the top-right corner of Figure 1. This environment featured a tranquil scene of a blue sky and swaying meadow, created using Unity Engine v2019. We adjusted the dynamics of the swaying meadow according to specific experimental conditions.

A 32-channel LiveAmp EEG system (Brain Products GmbH) was used to collect the EEG data at a sampling rate of 1000 Hz and a resolution of 40.7 nV. We mounted the electrodes on EasyCap (Brain Products GmbH), the layout of which followed the international 10/20 electrode placement system [66]. The electrode-skin impedance was maintained under 10 kΩ with an abrasive electrode paste (Abralyt HiCl). EEG data on all channels were acquired using FCz as the reference and were filtered using a band-pass filter of 0.0159 Hz and 250 Hz plus a 60-Hz notch filter.

The ArtiseBio Cygnus-819997-RawEEG system [67] was employed to collect ECG data at a sampling rate of 1000 Hz. Three electrodes from the ArtiseBio system were affixed to Kendall 200 series electrode pads (Cardinal Health) and positioned according to the standard limb leads II configuration for ECG monitoring [68].

A Vernier Go Direct Respiration Belt monitored and captured the respiration data at a sampling rate of 1000 Hz. The respiration belt was worn on the rib cage. After the initiation of each experimental condition (ie, each breathing technique), the program collected 10 seconds of a respiratory signal from each participant for normalization. We compared the respiratory signal with the normalized signal to determine whether the participant was inhaling or exhaling. The participant’s respiration was visualized in the VR environment throughout the experiment, except for the guided breathing condition.

The bottom-left corner of Figure 1 presents the synchronized EEG, ECG, and respiratory signals captured during the experiment, which were synchronized and coordinated using the Lab Streaming Layer. Multimedia Appendix 1 shows a screenshot of the synchronized data collection during an experimental session.

This study was a cross-sectional experiment involving 51 healthy young adults in a single group, and all participants finished all 4 experimental tasks. As such, this study did not have a randomized allocation of participants. The experiment sessions took place in 2 laboratories at 2 universities in northern Taiwan.

We designed the experimental procedure after consultation with 2 certified mindfulness-based stress reduction program instructors. The professional advice suggested that the sensation of breathing may relate to breezing. Therefore, we designed the swaying meadow in a VR environment to visualize a breeze and used the meadow to provide feedback for the participants’ breathing. The participants were new to breathing training, and we did not require them to meet the low respiration rate suggested by the previous mindfulness or breathing training practice guides [34,35,69-71]. The experimental design intentionally instructed the participants to provide their natural response.

The bottom-right corner of Figure 1 provides a visual representation of the experimental procedure. The experiment manipulators briefed the participants on the experimental procedure and assisted them throughout an experimental session in a face-to-face manner. Participants began by spending 5 minutes in a VR environment, immersing themselves in a setting with a tranquil blue sky and a meadow that swayed in response to their respiratory signals. The first 5 minutes was a warm-up session, and no data were recorded yet. Subsequently, the experiment progressed through 4 unique conditions, each lasting 6 minutes. Throughout these conditions, we meticulously recorded EEG, ECG, and respiratory signals. The instruction of each condition was given to the participants verbally by the experiment manipulators before a condition started.

This study used repeated measures ANOVA (rmANOVA) to examine differences in EEG band power for all channels, inflow for sink channels, outflow for source channels, and connectivity patterns among brain regions in the resting, mindful breathing, guided breathing, and breath counting conditions. Post-hoc t tests were used to further assess pair-wise differences among these conditions. We adjusted all statistical analyses involving multiple comparisons using the false discovery rate (FDR) [94].

Figure 2 shows the results of the rmANOVA and post-hoc analyses of band powers across the 4 experimental conditions, and Table 2 shows the result summary of EEG band power and EEG effective connectivity outflow and inflow analyses.

For the delta band, no significant differences in band power were detected in any region of the scalp among the 4 experimental conditions (Figure 2 [part 1a]; P values >.05).

For the theta band, rmANOVA revealed differences in EEG power specifically in the occipital region (Figure 2 [part 2a]; P values <.05). Particularly, post-hoc t tests showed that relative to resting, the conditions of mindful breathing, guided breathing, and breath counting led to decreased theta power in the occipital region (Figure 2 [parts 2b, 2c, and 2d]; P values <.05). Additionally, breath counting showed elevated theta power compared with mindful breathing and guided breathing in the left occipital region (Figure 2 [parts 2f and 2g]; P values <.05).

We found significant global band power differences in the alpha, low-beta, high-beta, and gamma frequency bands (Figure 2 [parts 3a, 4a, 5a, and 6a]; P values <.05). In the alpha band, mindful breathing did not exhibit any band power difference when compared with resting (Figure 2 [part 3b]; P values >.05). Conversely, guided breathing displayed lower band power than resting in both the frontal and bilateral parietal regions (Figure 2 [part 3c]; P values <.05). In contrast, breath counting led to elevated global band power compared with resting (Figure 2 [part 3d]; P values <.05). Guided breathing exhibited lower band power in the frontal, left temporal, and bilateral parietal regions than mindful breathing (Figure 2 [part 3e]; P values <.05). Breath counting, in contrast, demonstrated elevated band power across the entire scalp when compared with both mindful breathing and guided breathing (Figure 2 [parts 3f and 3g]; P values <.05).

The results of low-beta and high-beta bands were similar (Figure 2 [parts 4a and 5a]). Mindful breathing led to elevated band power compared with resting, but only in channel C3 (Figure 2 [parts 4b and 5b]; P values <.05). Guided breathing caused reduced band power across the entire scalp relative to both mindful breathing and resting (Figure 2 [parts 4c, 4e, 5c, and 5e]; P values <.05). Breath counting displayed lower band power in the frontal and bilateral regions than mindful breathing and resting (Figure 2 [parts 4d, 4f, 5d, and 5f]; P values <.05). It was noted that breath counting showed decreased low-beta band power levels in midline channels (Cz, Pz, and P3) relative to mindful breathing (Figure 2 [part 4f]; P values <.05), a trend not observed in the high-beta band (Figure 2 [part 5f]; P values >.05). Compared with guided breathing, breath counting exhibited elevated power in both the low-beta and high-beta bands across the central, parietal, occipital, and bilateral temporal regions (Figure 2 [parts 4g and 5g]; P values <.05).

For the gamma band, mindful breathing showed higher band power than resting only in channel C3 (Figure 2 [part 6b]; P values <.05). Both guided breathing and breath counting had lower band power globally than resting (Figure 2 [parts 6c and 6d]; P values <.05) and mindful breathing (Figure 2 [parts 6e and 6f]; P values <.05). Lastly, there was no band power difference between guided breathing and breath counting (Figure 2 [part 6g]; P values >.05).

Figure 3 shows the results of rmANOVA and post-hoc analyses of connectivity outflow among the 4 conditions. There was no connectivity outflow difference in the theta, low-beta, or high-beta bands in any of the scalp regions among the 4 conditions (Figure 3 [parts 2a, 4a, and 5a]; P values >.05).

For the delta band, rmANOVA showed significant intercondition differences in the left frontal and left parietal regions (Figure 3 [part 1a]; P values <.05). Mindful breathing and guided breathing showed no connectivity outflow difference when compared with resting (Figure 3 [parts 1b and 1c]; P values >.05). Breath counting led to lower outflow in the left frontal and left parietal regions than all the other conditions (Figure 3 [parts 1d, 1f, and 1g]; P values <.05). Guided breathing caused higher outflow in the left parietal region than mindful breathing (Figure 3 [part 1e]; P values <.05).

For the alpha band, rmANOVA showed intercondition differences in the right frontal, bilateral temporal, and left parietal regions (Figure 3 [part 3a]; P values <.05). Mindful breathing showed no connectivity outflow difference when compared with resting (Figure 3 [part 3a]; P values >.05). Both guided breathing and breath counting led to higher outflow than resting in the right frontal and central-parietal midline regions (Figure 3 [parts 3c and 3d]; P values <.05). Guided breathing led to higher outflow than mindful breathing in the right frontal region (channel Fp2; Figure 3 [part 3e]; P values <.05). Breath counting caused higher outflow than mindful breathing in the right temporal, central, and left parietal regions (Figure 3 [part 3f]; P values <.05), as well as higher outflow than guided breathing in the bilateral temporal and left parietal regions (Figure 3 [part 3g]; P values <.05).

For the gamma band, rmANOVA showed intercondition differences in the occipital and right parietal regions (Figure 3 [part 6a]; P values <.05). Mindful breathing and guided breathing displayed no connectivity outflow difference when compared with resting (Figure 3 [parts 6b and 6c]; P values >.05). Breath counting led to lower outflow than resting, mindful breathing, and guided breathing in the occipital and right parietal regions (Figure 3 [parts 6d, 6f, and 6g]; P values <.05).

Figure 4 shows the results of rmANOVA and post-hoc analyses of connectivity inflow among the 4 conditions.

For the delta band, differences in inflow across the 4 conditions were primarily observed in the bilateral frontal, right temporal, bilateral parietal, and occipital regions (Figure 4 [part 1a]; P values <.05). Mindful breathing exhibited reduced connectivity inflow in the left frontal region compared with resting (Figure 4 [part 1b]; P values <.05). While guided breathing showed no inflow difference relative to resting (Figure 4 [part 1b]; P values >.05), it did display increased inflow in the left occipital region relative to mindful breathing (channel O1; Figure 4 [part 1e]; P values <.05). Breath counting showed decreased inflow in the bilateral frontal, right temporal, bilateral parietal, and occipital regions relative to resting, mindful breathing, and guided breathing (Figure 4 [parts 1d, 1f, and 1g]; P values <.05).

For the theta band, only channel Fp2 showed an intercondition difference in the rmANOVA (Figure 4 [part 2a]; P values <.05). There was no inflow difference between the mindful breathing and resting conditions (Figure 4 [part 2b]; P values >.05). Both guided breathing and breath counting exhibited increased inflow in the right frontal region (channel Fp2) relative to resting (Figure 4 [parts 2c and 2d]; P values <.05). Guided breathing displayed enhanced inflow in the same right frontal region (channel Fp2) relative to breath counting (Figure 4 [part 2e]; P values <.05). There was no inflow difference between breath counting and both guided breathing and mindful breathing (Figure 4 [parts 2f and 2g]; P values >.05).

The alpha band showed differences in inflow across the entire scalp between conditions (Figure 4 [part 3a]; P values <.05). Mindful breathing led to increased inflow in the right frontal region (channel Fp2) compared with resting (Figure 4 [part 3b]; P values <.05). Guided breathing had a stronger inflow than resting in the frontal, left and midline parietal, and right central regions (Figure 4 [part 3c]; P values <.05). Breath counting exhibited increased inflow globally relative to resting (Figure 4 [part 3d]; P values <.05). Meanwhile, guided breathing showed increased inflow relative to mindful breathing in the right central and right parietal regions (Figure 4 [part 3e]; P values <.05). When compared with mindful breathing, breath counting led to greater inflow across all channels, except for Fp2, Fz, F3, and C3 (Figure 4 [part 3f]; P values <.05). When compared with guided breathing, breath counting caused enhanced inflow in the bilateral frontal, bilateral temporal, midline central, bilateral parietal, and right occipital regions (Figure 4 [part 3g]; P values <.05).

The rmANOVA revealed an intercondition difference in the low-beta band in the frontal region (Figure 4 [part 4a]; P values <.05). Both mindful breathing and guided breathing exhibited increased inflow compared to resting, specifically in the right frontal region (channel Fp2; Figure 4 [parts 4b and 4c]; P values <.05). There was no significant inflow difference between mindful breathing and guided breathing (Figure 4 [part 4e]; P values >.05). Breath counting showed decreased inflow in the midline frontal region relative to both mindful breathing and guided breathing (Figure 4 [parts 4f and 4g]; P values <.05), but it was comparable to resting (Figure 4 [part 4d]; P values >.05).

In the high-beta band, both mindful breathing and guided breathing demonstrated increased inflow in the right frontal region compared with resting (Figure 4 [parts 5b and 5c]; P values <.05). There was no observed inflow difference between mindful breathing and guided breathing (Figure 4 [part 5e]; P values >.05). Breath counting exhibited decreased inflow in the left frontal, right central, midline and right parietal, right temporal, and occipital regions in comparison with the other conditions (Figure 4 [parts 5d, 5f, and 5g]; P values <.05).

In the gamma band, mindful breathing displayed increased inflow in the left frontal region (channel F7) compared with resting (Figure 4 [part 6b]; P values <.05). Guided breathing showed elevated inflow in the right occipital region (channel O2) relative to resting (Figure 4 [part 6c]; P values <.05). No inflow difference was observed between mindful breathing and guided breathing (Figure 4 [part 6e]; P values >.05). Breath counting revealed decreased inflow in the left frontal, right central, and midline parietal regions relative to resting (Figure 4 [part 6d]; P values <.05). Furthermore, breath counting exhibited diminished inflow in the left frontal, midline and right central, midline and left parietal, and right occipital regions when compared with both mindful breathing and guided breathing (Figure 4 [parts 6f and 6g]; P values <.05).

Figure 5 shows the results of the rmANOVA and post-hoc analyses examining the connectivity flow among the 6 brain regions.

Notably, the delta, high-beta, and gamma frequency bands did not display any significant differences in connectivity flow across the 6 regions among the 4 experimental conditions (Figure 5 [parts 1a, 5a, and 6a]; P values >.05).

For the theta band, compared with mindful breathing, breath counting led to increased connectivity flow from the right frontal region to the central, parietal, and bilateral temporal regions; from the central region to the right temporal region; between the parietal and bilateral temporal regions; and from the left temporal region to the parietal region (Figure 5 [part 2f]; P values <.05).

The alpha band showed robust connectivity differences across conditions (Figure 5 [part 3a]; P values <.05). When compared with all the other conditions, breath counting showed greater connectivity flow from the right frontal to the central, parietal, and bilateral temporal regions (Figure 5 [parts 3d, 3f, and 3g]; P values <.05). Breath counting also had greater connectivity flow from the central region to all the other brain regions, as well as from the parietal region to the left frontal, central, and bilateral temporal regions (Figure 5 [parts 3d, 3f, and 3g]; P values <.05). Breath counting also had greater connectivity flow from the left temporal region to the left frontal and parietal regions, as well as from the right temporal region to the left temporal and parietal regions (Figure 5 [parts 3d, 3f, and 3g]; P values <.05). Compared with resting, guided breathing induced greater connectivity flow from the right frontal region to the central and right temporal regions and from the central region to the bilateral temporal regions (Figure 5 [parts 3d, 3f, and 3g]; P values <.05), as well as from the left temporal region to the central region (Figure 5 [part 3c]; P values <.05). Compared with mindful breathing, guided breathing had greater connectivity flow from the right frontal region to the central, parietal, and right temporal regions (Figure 5 [part 3e]; P values <.05); from the central region to the bilateral temporal and parietal regions (Figure 5 [part 3e]; P values <.05); and from the right temporal region to the parietal region and vice versa (Figure 5 [part 3e]; P values <.05). Overall, the right frontal region, but not the left frontal region, was the source of influence for alpha band connectivity in guided breathing and breath counting.

For the low-beta band, breath counting induced greater connectivity flow from the right frontal region to the left temporal region than the other conditions (Figure 5 [parts 3d, 3f, and 3g]; P values <.05).

We have presented the results and details of questionnaires in Table 3 and Multimedia Appendix 2, respectively. The results suggested that the participants had no extraordinary characteristics. The participants had medium mindfulness levels (FFMQ scores: mean 120.9, SD 15.7), medium sleep quality levels (PSQI scores: mean 5.5, SD 3.2), low depression levels (BDI scores: mean 12.6, SD 8.8), and low anxiety levels (BAI scores: mean 12.6, SD 8.8). As questionnaire responses are not the focus of this paper, we present the data in Table 3 without further interpretation.

Our research questions are whether and how people’s neurological features change when they are performing different breathing training techniques. The rmANOVA results revealed significant differences in EEG band power across the theta, alpha, low-beta, high-beta, and gamma bands over the entire scalp among resting, mindful breathing, guided breathing, and breath counting conditions. Outflow analysis identified condition-specific variations in the delta, alpha, and gamma bands, while inflow analysis showed significant variations across all frequency bands. Connectivity flow analyses showed that the right frontal, central, and occipital brain regions were predominantly influential in guided breathing and breath counting. Notably, in all the analyses, guided breathing and breath counting exhibited distinguishable EEG features compared to resting. In contrast, the EEG features of mindful breathing did not significantly differ from those of resting.

There were some limitations. First, because of the cross-sectional nature of this study, electrophysiological changes occurring during long-term breathing training could not be evaluated. Future investigations are essential to delve into the behavioral and neural transformations, and the underlying mechanisms induced by different VR-guided breathing training styles. This will contribute to a more comprehensive understanding of the long-term impacts and benefits of these interventions.

Second, as this study focused on novice mindful breathing practitioners, the evidence may not address the situation of how experienced breathing training practitioners would neurologically respond to VR-BF integrated training. It is also unclear whether VR-BF integrated technologies would make any difference (good or bad) to the training efficacy of experienced breathing training practitioners. Researchers may focus on experienced breathing training practitioners in terms of VR-BF integrated training.

Although the hardware of VR-BF self-help breathing training is relatively affordable, a missing piece for the popularization of such breathing training is the corresponding software. There is a need for an open-access breathing training application that integrates different models of VR headsets and BF devices (eg, smartphones, respiration belts, and consumer-level EEG headsets). This need is unaddressed. Subsequent studies may focus on the development of a VR-BF self-help breathing training application.

This study provides a list of electrophysiological features (ie, band power and effective connectivity) of different breathing training techniques. However, the long-term neurological mechanism changes due to different breathing training techniques are mostly unaddressed. Manufacturers can plausibly integrate their consumer-level EEG headsets with a breathing training application to monitor trainees’ electrophysiological signal changes. Such a breathing training program with EEG BF may then introduce the designated breathing training effect to trainees. We do not want to make any premature statement as the long-term electrophysiological features are yet to be discovered. Further studies on the EEG features of practicing different breathing training techniques for a longer period are needed for the commercialization of breathing training applications with EEG BF.

This study analyzed the data of 51 participants, and we believe that this size was sufficient to produce stable and reliable results. Previous empirical studies on VR-BF and breathing training usually had 9 to 45 participants [1,3,30,33], and the sample size of this study was relatively large in this area of research. To strengthen our analysis, we conducted FDR correction in all assessments [94]. Nevertheless, further studies with a larger sample size to examine the conclusion of this study are encouraged.

This preliminary study attempted to elaborate on the electrophysiological differences among different breathing training styles. Our results suggest that mindful breathing, guided breathing, and breath counting could be effectively administered through integrated VR and BF breathing training. We also observed the differences in EEG features among these breathing styles, particularly in band power and effective connectivity. These EEG features hold promise as potential biomarkers in integrated VR and BF breathing training for monitoring and assessing the progress of practitioners.