A total of 20 college students from the Shanghai Medical College of Health Sciences were selected for this study. The participants were 10 males (22.80 ± 0.81years) and 10 females (23.13 ± 0.92 years). The inclusion criteria are as follows: (1) Absence of any previous neurological or psychiatric diseases or other health problems; normal intelligence; (2) Age between 18 and 26 years (Sowell et al., 2003; Tamnes et al., 2010); (3) Right-handedness (Knecht et al., 2000; Beam et al., 2009); (4) Absence of contraindications to rTMS; (5) Informed consent was signed before the experiment. The exclusion criteria are as follows: (1) Administration of psychotropic drugs or neuroactive substances within the last 1 month; (2) intake of alcohol, caffeine >200 mg/day, or nicotine products 48 hours prior to enrolment; (3) presence of contraindications to rTMS (e.g., pacemakers, cochlear implants, and other metallic implants in vivo); (4) Female subjects are in pregnancy or on days 7–14 of the menstrual cycle (fluctuations of oestrogens during the luteal phase may affect fNIRS signals) (Xia et al., 2007; Chen D. Y. et al., 2023).

The prefrontal brain regions of the subjects were monitored using a 37-channel near-infrared functional brain imaging system, the BS-7000 (Wuhan Yiruide, China). The headcap probe is based on the 10–20 international standard lead design and consists of 12 transmitting optodes and 12 receiving optodes, which collectively constitute 37 acquisition channels. These channels are used to capture data from the subject's brain network. The distance between the probe and the light source is 30mm. The brain area below the midpoint of the probe and the light source is the primary probing area of the channel, which is calibrated concerning the Automated Anatomical Labeling (AAL). A template for labeling cerebral cortical subdivisions of the channel was employed, along with recording near-infrared light intensity signals at different wavelengths (690, 830 nm) at a sampling frequency of 100 Hz.

In this study, the primary observation area was the prefrontal lobe, which was covered by the fiber optic cap channel and divided into six brain regions according to the AAL brain regions. These were the bilateral superior frontal gyrus (SFG), middle frontal gyrus (MFG), and lower frontal gyrus (LFG). The results of the localization are shown in Figure 1.

The subject remained relaxed and wore a head cap after sitting quietly for 5 min in a comfortable position. During the scan, the subject remained relaxed with their eyes closed but did not fall asleep, and resting state data were collected.

This study employed a single-factor, six-level within-subjects design with two independent variables: intervention type—intermittent theta burst stimulation (iTBS)—and time point, which includes a baseline period (pre-TMS: T0) and five post-intervention time windows (post-TMS: 10, 25, 40, 55, and 70 min). Each complete module consists of a 30-s preparation period, a 5-min resting-state recording, and a 30-s recovery period. The blocking interval between modules is 10 min (Figure 2).

A YRD-CCY-I transcranial magnetic stimulation device (Iridium, Wuhan, China), with a 70-mm figure-of-eight coil and a maximum output magnetic field of 3.0 T, was employed to stimulate the DLPFC using a magnetic stimulation cap to guide coil placement. Before stimulation, the subject was comfortably positioned either supine or seated, with the entire body relaxed.

Before stimulation, the motor threshold (MT) was measured for each subject. The coil was placed in the left M1 area, tangential to the scalp, and the myoelectric amplifier was placed in the right thenar muscle to record the motor evoked potentials, with the anode affixed proximally, the cathode distally, and the ground wire affixed to the radial styloid process. Subjects exhibiting a relaxed state of the target muscles were administered a minimum of 10 consecutive single-pulse stimuli using TMS to ascertain the resting motor threshold (rMT). The rMT was defined as the minimum stimulus intensity that excites a wave amplitude >50 μV MEP at least five out of 10 stimuli, expressed as a percentage of the maximum output intensity of the stimulator.

Place the figure-eight coil tangentially on the skull surface, position its center over the left DLPFC, and set the stimulation intensity to 80% RMT. iTBS stimulation parameters: intra-cluster frequency 50 Hz, number of intra-clusters 3; inter-cluster frequency 5 Hz, number of inter-clusters 10; stimulation time 2s, interval time 8s; number of pulses 600; treatment time 200s.

The raw data of 37 channels were preprocessed using Matlab software running the Homer2 toolbox: (1) the raw light intensity was converted to optical density data; (2) motion artifacts were identified to reduce the disturbing factors generated by body micro-movements; (3) motion artifacts were removed by bar interpolation to patch the data; (4) the interfering signals from external and self-factors were removed by band-pass filtering from 0.01 to 0.10 Hz; (5) according to the modified Lambert–Beerlaw, the optical density data were converted into oxygenated hemoglobin (Oxy-Hb) concentration change values with a high signal-to-noise ratio; (6) the hemodynamic data of each block were superimposed and averaged. At the same time, we implemented strict signal separation techniques during preprocessing of fNIRS resting-state data. First, we used short-distance channels sensitive to superficial tissues to extract superficial hemodynamic signals. Then, we regressed these signals out of the target signals as noise sources to reduce contamination from extracranial components. Additionally, leveraging the characteristic that short light source-detector distance channels primarily reflect superficial tissue signals, we applied algorithms to correct long-distance channels, thereby isolating purer intracerebral hemodynamic information. These signal separation and correction techniques effectively enhance the signal-to-noise ratio of resting-state fNIRS data, providing a more reliable foundation for subsequent analysis. Subsequently, the data were subjected to analysis using the NIRS-KIT toolbox. Given that the signal-to-noise ratio of HbO₂ is superior to that of HbR, only the obtained HbO₂ data were employed in the subsequent data analysis (Cui et al., 2010; Wang and Lu, 2022; Luo et al., 2024).

Brain activation analysis: We performed task-like activation analysis on resting-state data to obtain brain region activation maps for each time point, based on previous literature (Chen D. Y. et al., 2023). A general linear model (GLM) was established for HbO₂ change values, and the resulting activation coefficient β values represent changes in cortical neuron activity within the corresponding channel. The larger the β value, the higher the activation level.

Brain network functional connectivity strength analysis: all resting state FC analyses were performed using NIRS-KIT. Pearson correlation coefficients were calculated between all channel time series to determine FC between each pair of measurement channels. This process yielded a 37 × 37 connectivity matrix for each patient. BrainNet Viewer software was used to plot a 3D map of the brain functional network based on the Pearson correlation coefficient matrix. Additionally, a circle plot was generated using a Matlab script based on the same matrix.

This experiment utilized GraphPad 9.0 for data processing and analysis. The Shapiro-Wilk test was employed to assess data normality. Quantitative data with normal distribution are expressed as mean ± standard deviation (x ± s). Paired-sample t-tests or one-way analysis of variance (ANOVA) are conducted accordingly. A p-value < 0.05 was considered statistically significant. Brain network visualization images were created using the Brain Connectivity Toolbox in MATLAB. Multiple comparisons of β values, functional connectivity (FC), and Pearson correlation coefficients were corrected using the false discovery rate (FDR) method for all statistical tests.

Functional near-infrared spectroscopy (fNIRS) was used to obtain blood oxygen activation heatmaps of the prefrontal cortex. In these heatmaps, blue-cool tones indicate low activation, while red-warm tones indicate high activation. Overall, the participants' prefrontal cortex showed different levels of activation at various time points. These times were both before and after iTBS stimulation. Compared to pre-stimulation (T0), the red-activated regions were more concentrated and larger in area at 10 min post-stimulation (T10), suggesting that iTBS temporarily enhanced task-related activation in the prefrontal cortex. Over time, the effects of iTBS gradually diminished: at T25 and T40, the red activation areas remained large, while at T55 and T70, these areas significantly decreased or even disappeared (Figure 3). A one-way analysis of variance was conducted to compare HbO₂ β values across all channels and time periods (Table 1). The results showed significant differences in channels 4, 15, 18, 22, 24, 25, and 28.

To observe differences in FC in the prefrontal cortex during resting state before and after TMS intervention, we used Pearson correlation coefficients to calculate correlations between channels. The resulting FC matrix was shown in Figure 4. The FC matrix directly reflected the connection strength between channels in the monitored brain region. The 37 × 37 matrix represents connection strengths among 37 channels. Because the FC matrix is symmetric, we analyzed only the lower triangular part, excluding the diagonal elements (self-connections) and the redundant upper triangular portion. Figure 4 showed that at different time points before and after iTBS intervention, FC intensity was strong in the bilateral superior frontal gyrus (SFG), middle frontal gyrus (MFG), and right inferior frontal gyrus (IFG). In contrast, FC intensity between the left IFG (IFG-L) and other brain regions was relatively weak.

To further investigate differences in FC in the prefrontal cortex at different time points before and after iTBS stimulation, we first selected detection channels with statistically significant differences. Then, we plotted 3D images of brain network FC between these channels and their corresponding brain regions (Figure 5A), as well as circle plots (Figure 5B). Based on the observation of channel coverage locations, FC was significantly enhanced during the T10 period compared to T0, both between key channels and their corresponding brain regions. Statistical results for FC intensity (Figure 6) showed that FC intensity was significantly elevated in the early stage (T10) following a single iTBS intervention compared to baseline (T0). However, this effect clearly decayed over time: there was no difference between the T55 and T0 periods, and FC intensity returned to baseline levels by T55. Although FC intensity in the T70 period was lower than in the T55 period, there was no statistically significant difference between the two. These results confirm that the neuromodulatory effect of a single iTBS intervention lasts up to 40 min, providing a key time window for optimizing clinical treatment intervals.