Healthy adults without any known cardiovascular or neurological disorders, voluntarily recruited through public advertisements. Sixty-one healthy participants (30 males and 31 females, mean age: 22.21 ± 3.15 years) were enrolled in the study. The sample size for each group was determined with reference to previous neuromodulation studies using TN stimulation [19, 20]. Sixteen participants were assigned to the sham group, and 15 participants each to the 2, 20, and 200-Hz stimulation groups, yielding a total sample of 61.

Participants were randomly assigned to one of the stimulation frequency groups (sham, 2 Hz, 20 Hz, and 200 Hz) in a single-blind design, such that they were unaware of their group allocation. Two participants in the sham group were excluded from the final analysis due to data loss caused by measurement equipment errors. The flow diagram illustrating participant recruitment and inclusion in the final analysis is presented in Fig. 1, and the demographic characteristics of each group are summarized in Table 1.

Flow diagram of participant recruitment and inclusion in analysis

Participants were instructed to abstain from drinking alcohol, smoking, and consuming caffeine the day prior to the experiment. Individuals with cardiovascular disease were excluded from participation. The experimental procedures were reviewed and approved by the Institutional Review Board of Hanyang University (Approval No. HYU-2020-03-010). All participants provided written informed consent and received a small financial incentive for their participation.

The experiments were conducted in the late morning (between 9:00 AM and 12:00 PM). Before the experiment began, participants were attached with Ag/AgCl electrodes (Kendall 100 series foam electrodes, Cardinal Health, Inc., USA) for electrocardiography (ECG) measurement at four locations on the chest and abdomen (Fig. 2B). Photoplethysmography (PPG) sensor was attached to the tip of the finger. To minimize changes in autonomic nervous system activity, participants were instructed to maintain a respiratory rate of 20 breaths per minute, synchronized with a metronome (FM-310, Fzone) and the respiration signal was monitored by band-type respiration sensor (TSD201, Biopac System Inc, USA). To reduce extraneous noise, participants wore wireless noise-canceling earphones (AirPods Pro, Apple).

Overview of experimental setup and data acquisition. A The stimulation sites and parameters. The stimulation was applied to the ophthalmic and maxillary branches of the TN on the face. B The locations for bio-signal data acquisition, where blue indicates ECG measurement sites, yellow represents respiration monitoring sites, and orange denotes PPG measurement locations. C The experimental protocol, with the blue lines indicating rest periods and the yellow lines representing stimulation periods

The experimental protocol, as depicted in Fig. 2C, lasted 21 min. It began with a 7-min resting baseline, followed by alternating 2-min sections with electrical stimulation OFF and ON, repeated three times. The protocol concluded with a final 2-min OFF period. During the experiment, ECG, PPG, and respiration signals were continuously recorded in real time. ECG and PPG signals were sampled at 1000 Hz using the Taskforce Monitor (Taskforce Monitor 3040i, Austria), whereas respiration signals were sampled at 2000 Hz using the Biopac system (MP160, Biopac Systems Inc., USA). This setup was necessary because the Taskforce Monitor does not include a respiration measurement module; thus, a separate Biopac system was employed for monitoring respiration. The Biopac system is configured to acquire high-resolution signals by default, which resulted in the respiration signal being recorded at a higher sampling frequency. In this study, respiration signals were used to confirm that participants were following the instructed breathing protocol; therefore, they were not included in the analysis or presented in the results.

Medical low-frequency stimulation pads (Nu Eyne Co., Ltd., Korea) were placed on the forehead, just above the eyebrows, and on the cheeks, just below the eyes. These pads were specifically designed to target the ophthalmic and maxillary branches of the TN. Electrical stimulation was delivered using Nu Eyne’s clinical trial stimulator (TPD-NH1, Nu Eyne Co., Ltd., Korea) in biphasic pulses with a phase duration of 250 µs. The stimulation frequencies were set to 2 Hz, 20 Hz, and 200 Hz, depending on the group (Fig. 2A). Participants adjusted the stimulation intensity themselves to below level where they experienced mild discomfort. The average stimulation intensities were as follows: 2 Hz group: 3.39 ± 1.11 mA, 20 Hz group: 2.99 ± 1.11 mA, and 200 Hz group: 2.60 ± 0.80 mA.

All bio-signal data were processed using MATLAB software (MathWorks, USA). First, the R–R interval (RRI) was determined by identifying the R peaks in the ECG signal. Continuous HR was then calculated by dividing each RRI into 60 s after resampling. The calculated HR values for each group were averaged and smoothed using a moving median filter with a window length of 5 s.

Feature points from the PPG signal and the RRI from the ECG were extracted to calculate the PAT, defined as the time difference between the R peak of the ECG and the corresponding feature point in the PPG. The feature point in the PPG was identified as the maximum peak of the first derivative of the PPG signal. After resampling, the PAT values were averaged and filtered with a moving median filter using a 5-s window.

To evaluate changes in cardiovascular signals between stimulation ON and OFF phases, a normal probability density distribution was constructed using the mean and standard deviation of the preceding OFF phase. This distribution was then used as a reference to quantify how much each data point in the ON phase deviated from the baseline variability. A two-tailed test with a significance level of 0.05 was applied, and Bonferroni correction was implemented to adjust for multiple comparisons. This analytical procedure was applied consistently to both HR and PAT datasets. All statistical analyses were performed using functions from MATLAB’s Statistics and Machine Learning Toolbox.

Autonomic nervous system (ANS) indicators were analyzed using the HRVAS toolbox, a MATLAB add-on. HRV analysis was conducted across all TN stimulation groups (2, 20, and 200 Hz; sham group was excluded; N = 45). The high-frequency (HF, 0.15–0.40 Hz) band of HRV was used as a marker of parasympathetic activity, while the low-frequency (LF, 0.04–0.15 Hz) band represented sympathetic activity. To account for the fixed respiratory rate during the experiment (20 breaths per minute, equivalent to 0.33 Hz), the HF band was adjusted to 0.15–0.3 Hz. ANS balance was quantified using the LF/HF ratio, which reflects the relative contributions of sympathetic and parasympathetic activity. To further evaluate the effects of TN stimulation on autonomic balance, the correlation analysis was performed to examine the relationship between the LF/HF ratio in the rest phase and the change in LF/HF ratio during the first stimulation ON phase. Prior to this analysis, the normality of both variables was assessed using the Shapiro–Wilk test, which indicated that neither variable followed a normal distribution (p < 0.05). Consequently, Spearman’s rank correlation, a non-parametric method, was applied to assess the association between these variables.