For 7T MRI scanning, we recruited 20 student table tennis players (aged 22.2 ± 2.75 years, 12 men, and 15.6 ± 2.11 years of education) and 21 healthy control participants who were matched in gender, age, and educational level (aged 22.62 ± 2.41 years, 12 men, and 16.62 ± 2.15 years of education; Table 1). All student athletes were Chinese national first-class table tennis players (“Athlete Technical Grade Standard” issued by the General Administration of Sports of China was considered the evaluation principle) with >6 years of training experience, whereas none of the healthy control participants had any sports specialties or hobbies. All the participants exhibited normal or corrected to normal vision and were right-handed with no history of neurological and psychiatric diseases. All experimental procedures were approved by the research ethics review committee of Wuyunshan Hospital ([2021] Clinical Research Medical Ethics Review No. 004) and conducted in accordance with the Declaration of Helsinki. All the participants provided informed written consent before the experiment and no drugs were administered during the experiment that could alter mental status, cognition, or perception.

MRI scanning experiments were performed using Nova Medical 32-channel array head coils in a 7T whole-body magnetic resonance system (Siemens Healthcare, Erlangen, Germany) and included rs-fMRI and structural image scanning. rs-fMRI data were acquired using an echo-planar imaging sequence with 1.5-mm isotropic resolution with the following parameters: repetition time (TR) = 2,000 ms, echo time (TE) = 20.6 ms, flip angle = 70°, field of view (FOV) = 228 mm × 228 mm, number of slices = 90, and total volume = 160. Structural images were acquired using the MP2RAGE sequence with 0.7- mm isotropic resolution with the following parameters: TR = 2,300 ms, TE = 3.09 ms, flip angle = 7°, and FOV = 225 mm × 225 mm. During the scan, the participants were asked to close their eyes, remain awake, and stay as still as possible.

Resting-state functional images were preprocessed using RESTplus V1.24¹ (Jia et al., 2019) based on MATLAB 2017a (Mathworks, Natick, MA, United States).

The first five time points of resting-state fMRI data were discarded owing to the brief adaptation period required when a participant first enters the scanner and the instability of the initial MRI signal, leaving 155 time points per participant for further analysis (Liu et al., 2022). Head motion correction was performed, and the participant was excluded if their head movement exceeded 2.5 mm in any direction or 2.5° in any angle (Yan et al., 2013). Individual structural images were coregistered with the mean functional images and segmented into gray matter, white matter, cerebrospinal fluid, skull, extracerebral, and soft tissue using NewSegment tool in RESTplus V1.24. Coregistered structural images were normalized to the standard Montreal Neurological Institute space. For subsequent statistical analysis, a Gaussian kernel with 6 mm full width at half height (FWHM) was used for spatial smoothing; the linear trend of the time series was subsequently eliminated. To reduce the effect of physiological noise, the following interfering covariates were regressed: 24 head motion parameters, white matter signal, and cerebrospinal fluid signal (Friston et al., 1996). Finally, the FC was calculated after temporal band-pass filtering (0.01 to 0.08 Hz). Note that ALFF was calculated without filtering during preprocessing. No participants were excluded from the final dataset for analysis.

Three frequency bands (conventional frequency, slow-4, and slow-5 bands at 0.01–0.08, 0.027–0.073, and 0.01–0.027 Hz, respectively) of sALFF values were calculated using RESTplus V1.24. Specifically, the preprocessed time series was converted to a frequency domain using a fast Fourier transform, and the power spectrum was acquired. Average values were then calculated (average value of the square root of the power) at different amplitudes to obtain the ALFF values. Finally, the ALFF of each voxel was divided by the global average ALFF value to achieve standardization (Zang et al., 2007). The sliding window method was used to evaluate the dALFF using the temporal dynamic analysis module of RESTplus V1.24 (Jia et al., 2019; Gao et al., 2021; Figure 1). According to recommendations, the minimum window length should be >1/fmin, where fmin indicates the minimum frequency of time courses (Leonardi and Van De Ville, 2015). Based on previous studies, we chose a rectangular window with a length of 50 TRs and a moving step of 1 TR, producing 106 segmented windows (He et al., 2019; Fateh et al., 2020). The ALFF map was then computed for each window. Finally, based on a series of ALFF maps in all the windows, the PerAF corresponding to each voxel was computed, producing a dALFF map (Li et al., 2019; Gao et al., 2022). To verify the reliability of the metric of PerAF, the CV of the dALFF plot across windows was also calculated (Gao et al., 2022; Supplementary Figure S1).

After calculating sALFF and dALFF, the groups were compared to identify several significantly differing brain regions. These were selected as seeds to perform voxel-wised sFC and dFC calculations. sFC and dFC values were calculated at the voxel level using RESTplus V1.24. When calculating sFC, the Person correlation coefficient (r value) was calculated between the average time series of the seed region and every voxel of the whole brain (Bandettini et al., 1993; Gao et al., 2019; Mitra et al., 2023; Shahsavarani et al., 2023; Thippabhotla et al., 2023). Next, a Fisher’s r to z translation was applied to convert the r value to a normalized z value. The whole-brain FC map of each seed region could then be obtained. To calculate dFC, the same sliding window parameters were used as for dALFF (Figure 1). For every sliding window, a whole-brain z values FC map could be obtained between the mean time series of the seed region and every voxel of the whole brain (Gao et al., 2022). Notably, the number of FC probability maps was equal to the number of sliding windows. Finally, based on a range of z values FC maps in all windows, the PerAF of each voxel was calculated to obtain a dFC map. To verify the reliability of the metric of PerAF, the CV of the dFC map across windows was also calculated (Gao et al., 2022; Supplementary Figure S2).

The PerAF is one index of percentage signal change that is used to measure the BOLD signal fluctuations during rs-fMRI (Jia et al., 2020). Herein, we used PerAF to measure the percentage of dALFF and dFC change by converging the dALFF and dFC values of all the windows. The PerAF of each voxel is calculated as:

During the calculation of dALFF, X_i refers to the ALFF value of the i_th window. During the calculation of dFC, X_i is the z value of the i_th window, and the z value is converted from the r value of the mean FC in i_th windows using Fisher’s r to z translation. n is the number of sliding windows, and μ is the average of ALFF or z values of the n sliding windows.

The experimental procedure was adapted from the Eye-Hand coordination paradigm of the Nike Sensory Station assessments and was written using Psychophysics Toolbox based on MATLAB (MathWorks, Natick, MA, United States; Erickson et al., 2011; Farshchian et al., 2018; Figure 2). Stimuli were shown on a linearized monitor (1920 × 1,080 resolution, 100-Hz refresh rate) and drawn against a gray (56 cd/m²) background (Liu et al., 2022). The participants stood with their feet shoulder-width apart at an arm’s length from the wall and 1 m from the horizontal screen. The researcher adjusted the height of the buttons on the wall to the level of the participant’s shoulders. The visuomotor stimulus comprised 10 trials; each trial included a waiting and pressing state, and the participants were asked to focus on the screen the entire time. In the waiting state, a black fixation point appeared at the center of the screen. At this time, the participants kept their arms hanging down naturally and close to their thighs. After 5 s, the black spots disappeared and a red box immediately appeared in the middle of the screen. The participants were asked to raise their hands and press the button on the wall when they saw the red box, and then immediately returned to the waiting state. This cycle was repeated 10 times. The participants could familiarize themselves with the experimental operation before the formal experiment started. The reaction time elapsed from the appearance of the red box until the participant pressed the button was recorded.

The demographic information of all the participants was statistically analyzed using GraphPad prism 8. Chi-square test was performed to compare sex, and two-sample t-test was used for to compare age, years of education, and mean frame-wise displacement (FD) parameters. The threshold of statistical significance was established at a level of p < 0.05.

Statistical analysis of the four kinds of rs-fMRI for the athlete and control groups was performed using RESTplus V1.24, and two-sample t-test was performed for the analysis. To avoid the effect of covariates, regression analysis was performed for age, gender, education, and body mass index. For the results of two-sample t-test using Gaussian random field (GRF) theory for multiple comparisons correction, the voxel level threshold was p < 0.01 and the cluster level threshold was p < 0.05. The magnitude d effect size and confidence intervals were calculated.

For behavioral data, we used mean ± 2SD to exclude outliers and then calculated the average individual reaction time as behavioral indicators. SPSS 20 (IBM, United States) was used to evaluate the correlation between the mean reaction time and MR variables (sALFF, sFC, dALFF, and dFC) via Pearson’s correlation analysis. The mean reaction time differences between the groups were calculated using two-sample t-test. Differences or correlations were considered statistically significant if p was <0.05. The magnitude d effect size and confidence intervals were calculated.

The demographic characteristics of the athlete and control groups are shown in Table 1. No significant differences in age (t = 0.6794, p = 0.5009), sex ( χ 2 = 0, p > 0.9999), BMI (t = 0.3060, p = 0.7612), education (t = 1.494, p = 0.1431), and mean FD (t = 0.3933, p = 0.6963) were observed between the athletes and controls.

We found that sALFF and dALFF provided complementary information when studying differences in brain activity between athletes and controls. Three frequency bands (conventional frequency, slow-4, and slow-5 bands at 0.01–0.08, 0.027–0.073, and 0.01–0.027 Hz, respectively) were selected to calculate sALFF. Two-sample t-test results revealed that athletes exhibited smaller sALFF values in the left inferior temporal gyrus (lITG) when compared with those in controls in the traditional and slow-4 frequency bands; the t-values of the two frequency bands were different (Table 2; Figures 3A,B; Supplementary Figure S3, GRF correction, voxel p < 0.01 cluster p < 0.05). However, significant differences were not present between the groups in the slow-5 frequency band.

In addition, we used PerAF to measure the dALFF to represent the BOLD signal fluctuations. Athletes had markedly higher dALFF values in the right postcentral gyrus (rPoCG) than those in the controls, whereas dALFF values in the right middle & paracingulate gyri (rMCC) were reduced in athletes (Table 2; Figure 3C; GRF correction, voxel p < 0.01 cluster p < 0.05). All effect size statistics are expressed as Cohen’s d, which were > 0.8, indicating that our results are stable and reliable (Seymour et al., 2016).

Three brain regions, lITG, rMCC, and rPoCG, were selected as seeds, which were used to perform whole-brain voxel-wised sFC and dFC calculations. The brain regions showing significantly different static and dynamic functional connections between the two groups are shown in Figure 4 or Supplementary Figure S4 and the results are summarized in Table 3 (GRF correction, voxel p < 0.01, cluster p < 0.05).

Group differences of sFC and dFC were only present when the lITG was used as a seed, whereas the rMCC and rPoCG seed regions did not yield significant results in sFC and dFC analysis. When the lITG was selected as the seed for sFC analysis, the athlete group exhibited a significantly decreased sFC in the right middle temporal gyrus (rMTG) and left inferior parietal gyrus (lIPG). From the perspective of dFC, athletes exhibited increased dFC from the lITG to the left middle frontal gyrus (lMFG), left superior frontal gyrus-medial (lSFGmedial), left superior frontal gyrus-dorsolateral (lSFG), and lIPG compared with that in controls. These results were corrected using GRF correction (voxel p < 0.01 and cluster p < 0.05). The effect sizes are expressed using Cohen’s d values, which were > 0.8, thus indicating that our results are stable and reliable.

The visuomotor reaction time of the athlete group was significantly shorter than that of the control group (mean ± 2SD, athletes: 0.5252 ± 0.1085, controls: 0.6336 ± 0.1414, t = 5.351, p < 0.0001, d effect size = 1.6719, confidence intervals = [0.9486, 2.3793], df = 39, Figure 5A). The signal of each participant of the rs-fMRI index was extracted separately, and the correlation with the individual average visuomotor reaction time was calculated. Abnormal values for both the rs-fMRI index and visuomotor reaction time were excluded if they fell outside the range of mean ± 2SD. Following this principle, we excluded the data of two athletes. This showed that the dALFF value in the rMCC brain region of the athletes was significantly positively correlated with the average visuomotor reaction time (r = 0.6403, p < 0.01), however, correlation was not present in the control group (Figure 5B). No other correlations were detected in the other brain regions.

Herein, the results of static and dynamic analyses were complementary and overlapping. Multiband sALFF and dALFF analyses revealed compared with the controls, the athletes exhibited a lower sALFF in the lITG and rMCC, whereas they had a higher dALFF in the rPoCG. We found that the dALFF of rMCC could predict the visuomotor reaction time of table tennis players. Regarding sFC and dFC analyses, differences were present between athletes and controls only when the lITG was used as the seed region. Significant differences were present in two brain regions in the sFC analysis and four brain regions in the dFC analysis. Both sFC and dFC analysis revealed changes in the FC between lITG and lIPG. In particular, the sFC values of the brain regions with group differences were smaller in athletes than in controls, whereas the dFC values in athletes were larger than those in controls (Gao et al., 2021). These findings suggest that the correlation between lITG and lIPG was weaker in athletes compared with that in controls at baseline, and that the fluctuations in their correlation coefficients over time were greater in athletes than in controls (Vahdat et al., 2011; Huang et al., 2018; Gao et al., 2021; Iraji et al., 2021; Zheng et al., 2021). Therefore, we speculate that the joint static and dynamic analysis can obtain complementary and overlapping results, enabling a more comprehensive exploration of brain activity than analysis by a single method. Notably, there are more dynamic results than static ones. We speculate two reasons for why dALFF and dFC outperformed sALFF and sFC in this study: first, it may be because we used the PerAF index when calculating the dynamic index, and the introduction of the percentage increases the sensitivity of the dynamic index (Chen et al., 2022; Gao et al., 2022). Second, dFC markers encode more behavioral information compared with static FC markers, such as processing speed and executive function (e.g., working memory and attention; Liégeois et al., 2019). Our dFC analysis detected changes in the FC of vision and frontal cortex brain regions, which are also key brain regions for executive function (Knöllner et al., 2022). Furthermore, the experimental data was obtained using 7T ultra-high field magnetic resonance, which has a higher temporal signal-to-noise ratio, and therefore 7T ultra-high field magnetic resonance may be suitable for the dynamic analysis of rs-fMRI (Hale et al., 2010). The above explanations remain speculation, and further experimentation of the same type is required to support this.

Static and dynamic joint analysis revealed that table tennis players exhibited alterations in intrinsic brain activity in visual and motor function-related brain regions, with decreased cortical activity being the mainstay. Our behavioral results revealed that table tennis players exhibited a faster visuomotor reaction time than non-athletes. Previous studies reported that superior athletic performance in elite athletes may be associated with lower activity in some brain areas compared with non-athletes, consistent with the neural efficiency hypothesis (Del Percio et al., 2009; Neubauer and Fink, 2009; Guo et al., 2017; Danti et al., 2018; Li and Smith, 2021; Gao et al., 2023). This is consistent with our results. lITG is a high-level brain area of the ventral visual pathway, which is related to the position, color, posture, and facial recognition of objects. lITG is a brain region that is arguably central for visual object recognition in both humans and nonhuman primates (Li and DiCarlo, 2008; Jia et al., 2021). In a fMRI study of table tennis players on a visuospatial task, Guo et al. found that athletes exhibited lower brain activation in the regions of the left MTG, bilateral lingual gyrus, and lITG than non-athletes (Guo et al., 2017). Table tennis requires extensive visual participation. Players need to quickly receive and process various visual information, such as tracking the trajectory of the ball and observing the opponent’s hand and racket posture (Rodrigues et al., 2002; Hülsdünker et al., 2019; Piras et al., 2019; Shinkai et al., 2022). After long-term training, these athletes have formed a unique visual strategy, with the lITG function probably becoming more specialized, which is conducive to rapid response with higher neural efficiency (Yin et al., 2021).

dALFF analysis revealed that table tennis players also exhibited significantly reduced activity of rMCC. In a broad sense, the cingulate cortex is also a part of the SMN, which includes the midcingulate cortex (MCC; Mitchell et al., 2013; Herz et al., 2016; Vogt, 2016). The motor function of the MCC is critical and is mainly concentrated in the cingulate premotor area at the tail of the posterior MCC, which is involved in visual and spatial location and multisensory orientation of the head and body in space (Braak, 1976; Matelli et al., 1991). The neuron responses of the MCC were tuned for the force and direction of movement (Braak, 1976; Matelli et al., 1991). Table tennis is a fast racket sport. In addition to players requiring rapid visual processing, they need to accurately perceive the direction and make rapid direction adjustments to the body to hit the ball quickly and accurately (Rodrigues et al., 2002; Kaabi et al., 2022; Takami et al., 2022). Existing evidence shows that MCC receives information projections from the parietal lobe, especially the posterior parietal cortex (PPC) to regulate body orientation, reflect movement, and make rapid motor responses, through coordination with the supplemental motor area and connections to the spinal cord (Vogt, 2016). We found that the dALFF value of rMCC in table tennis players was significantly positively correlated with the visuomotor reaction time. A lower value means a shorter reaction time, which also proves that rMCC can quickly regulate visual movement with high neural efficiency, enabling athletes to have better performance.

sALFF and dALFF analysis provided the functional separation results of the brain regions. To study the functional integration of brain regions, we calculated voxel-wised sFC and dFC and found that the athlete group exhibited significantly lower sFC in lIPG and rMTG. lIPG is a part of the PPC that lies between the sensory and motor areas and is a multisensory cortical area that receives input from multiple sensory systems (including somatosensory, auditory, visual, motor, proprioceptive and vestibular signals; Grefkes et al., 2004; Whitlock, 2017). The PPC is a key brain region for sensorimotor translation and is involved in motor planning (Li et al., 2022). Table tennis players must frequently initiate targeted motor responses without extensive cognitive processing, and visual information must be quickly translated into motor commands. Information flows from the visual area to the PPC and then to the premotor and supplementary motor areas to complete the visuomotor transformation (Hülsdünker et al., 2018). We found that sFC decreased between lITG and rMTG. Tankus et al. demonstrated that the temporal lobe plays a role in transforming visual information into hand movements, and this connection may be linked to the dorsal pathway (Tankus and Fried, 2012; Breault et al., 2019). lITG is a high-level brain region of the ventral visual pathway, whereas MTG is involved in the perception of tools, body and hands, movement observation, and visuomotor (Lesourd et al., 2021; Amaral et al., 2022). The MTG helps predict self-generated movements by optimizing them based on the physics of the end effector, such as the hand or tool being used (Pazen et al., 2020). Table tennis places great emphasis on observing and predicting hand and racket posture. Through extensive training, players develop an efficient process for transforming visual input into rapid visual motor responses, allowing the execution of precise hand movements with ease. Thus, we found that table tennis players exhibited reduced cortical activity during several key processes of visuomotor conversion, which was consistent with the neural efficiency hypothesis. After long-term professional learning and training in table tennis, athletes exhibited high levels of automaticity in visual motor responses. Bassett et al. found that decreased integration between motor and visual modules was observed during motor skill-learning (Bassett et al., 2015). Decreased ALFF or FC in visual and sensorimotor brain regions reduces energy consumption to ensure that athletes can perform with high neural efficiency (Bertollo et al., 2020; Filho et al., 2021).

Unlike the reduction in brain activity described above, table tennis players exhibited increased dFC when lITG was used as the seed region compared with controls. The observed functional changes in intrinsic brain activity in athletes via dFC analysis were primarily between the lITG and frontal cortex (lMFG, lSFGmedia, and lSFG). Traditionally, the frontal lobe is associated with executive function (Miyake et al., 2000; Cristofori et al., 2019). The middle frontal gyrus and superior frontal gyrus are two high-level cognitive brain areas located in the prefrontal cortex and involve executive functions, such as attention and working memory (du Boisgueheneuc et al., 2006; Li et al., 2013; Briggs et al., 2021). The lMFG is also considered a part of the ventrolateral prefrontal cortex (Kumar et al., 2022), which in primates receives visual information from the lITG (Romanski, 2007). The superior frontal gyrus includes SFGmedia and SFG, which are related to the cognitive control and default mode networks, and is strongly correlated with controlling and processing emotional/cognitive behavior (Li et al., 2013; Xu et al., 2022). In addition, the SFG plays an important role in the control and coordination of complex movements (Martino et al., 2011). In other studies involving athletes, changes in intrinsic brain activity were also found in the MFG, SFGmedia, and SFG. Elite snowboarding halfpipe athletes exhibited stronger activations in the MFG and SFG in response to incongruity when observing maneuvers; 8 weeks of Tai Chi Chuan exercise enhanced the functional connection between lMFG and the left superior parietal lobule; and soccer athletes exhibited increased brain activity in the bilateral MFG during motor control (Cui et al., 2019; Shi et al., 2020; Zhao et al., 2021). Visual information is critical to executive function. The lITG is a higher brain region of the ventral visual pathway (Knöllner et al., 2022). Increases in dFC from lITG to these three frontal brain regions (lMFG, lSFGmedia and lSFG) indicated dynamic instability and flexibility of the visual and executive areas in table tennis players. Therefore, the significant increase in dFC between lITG and the prefrontal cortex may be related to greater executive abilities, such as visuospatial working memory and visual selective attention in athletes (Huijgen et al., 2015; Wang et al., 2016; Elferink-Gemser et al., 2018).

Interestingly, regions with different cognitive and processing demands represent different levels of dFC. The brain networks/areas known to be involved in higher cognitive processing show a higher level of dynamism measured via dFC than that of networks/areas engaged in primary processing (Iraji et al., 2021). This may be a reason for the enhanced dFC between the lITG and prefrontal cortex.