The data presented here are part of a larger study of the effects of acute training on cognitive performance [40]. The focus of this article is to compare the changes in individual cortical activations during a fine motor-cognitive task (dTMT-B) in a seated position after a 10-minute post-exercise recovery period with two different types of exercise intervention. Our goal was to confirm the intervention effects of post-exercise recovery between sessions. Another previously published study examined the acute effects of two exercise protocols on hemodynamic response during dTMT while walking [40].

Sample size calculations were estimated using G*Power (Version 3.1.9.2) (Math.-Nat. Faculty, Düsseldorf, Nordrhein-Westfalen, Deutschland) [41] and were based on small to medium effects in prior studies of acute exercise in healthy adults [9]. A total sample size of 22 is required to detect a small to medium effect in a two-group design with a 5% risk of type 1 error ($\alpha$), 80% power, and an estimated correlation of r = 0.5 between repeated measurements of the cognitive outcome (dTMT-B). We recruited 32 participants to account for a dropout rate of $\sim$25%.

A total of 32 healthy young adults (7 women) between the ages of 19 and 33 participated in the present study. Inclusion criteria were right-handedness, at least 2 hours of exercise per week, and fair or good fitness (men: 43.5 mL/kg/min and above; women: 33.6 mL/kg/min and above), respectively [42]. Exclusion criteria were an abnormal resting pulse or blood pressure and cardiovascular or neurological diseases. Furthermore, participants should not have any fine motor impairments as this would hinder the performance of the cognitive task. Injuries that interfered with walking on the treadmill or scalp injuries that would interfere with measuring brain activity by fNIRS constituted another exclusion criterion. In order to recruit participants, our University Sports Club shared flyers, both manually and via e-mail, through which participation in the experiment was handled. Initially, 3 participants were excluded from the experiment, two for not reaching the required fitness level and one for the technical problem (failed recording for the fNIRS data). Thus, data from 29 participants were considered for the final analysis. The study was conducted in accordance with the Declaration of Helsinki and was approved by the university’s local ethics committee. We obtained written consent to participate in the experiment from all participants.

As a randomized controlled trial design using repeated measures (five sessions) in a two-armed parallel-group (between subjects), two laboratory visits were required for study participants (i.e., sessions (1) and (2); see Fig. 1), with an interval of at least two days between visits. During each visit, the experiment lasted approximately 1.5–2 hours. Before the experiment, participants received several instructions: They were not to exercise for 24 hours, not consume caffeine or alcoholic beverages for the last 12 hours, and sleep for approximately 7 hours the night before [43]. Fig. 1 shows the entire study procedure.

Fig. 1.

Schematic illustration of the design and testing procedures consisting of two test days for the HIIE and MCE groups. Cortical hemodynamic activation was measured using functional near-infrared spectroscopy (fNIRS) while participants performed the digital Trail-Making Test (seated or during running on a treadmill).

On the first day, after the participants gave their written consent to participate in the experiment, they completed a questionnaire on demographic information. Then the fNIRS system was prepared, and sources and detectors were attached to the head cap of the participants (see 3.3 fNIRS). After preparation, an initial resting value of 2 minutes was measured in a seated position. The Mental and Physical State and Trait Energy and Fatigue Scales Questionnaire questionnaire (only Part II; MPSTEFS) was then completed to determine participants’ physical and mental energy/fatigue [44]. The first dTMT session (session 1: on (1) of Fig. 1) followed while participants were comfortably seated at the table. After a 20-minute rest, the second dTMT session ((2) of Fig. 1) was performed again while sitting comfortably. Participants then completed the questionnaire in the MPSTEFS Part II again. At the end of the first day, we removed the fNIRS system and determined the individual fitness level (VO${}_2$max) by performing the Bruce protocol [45] on a treadmill (h/p/cosmos pulsar® 3p, Nussdorf-Traunstein, Germany). Participants were then randomized into one of two groups (HIIE and MCE).

On the second day, the cap with the sources and detectors of the fNIRS was placed on the head of the participants. An initial fNIRS measurement was recorded for 2 minutes at rest. Then answers to the MPSTEFS questionnaire were collected. Next, participants walked on the treadmill for 2 minutes at a speed of 5 km/h (including collecting data with fNIRS), followed by a dTMT session during walking (session 3: on (3) of Fig. 1). After completing the dTMT, the fiber optic cables of the optodes were disconnected from the fNIRS system and secured to the participants’ backs to complete the remainder of their exercise sessions without interruption (HIIE or MCE). At the end of the intervention, we reconnected the cable of the optodes to the fNIRS system. Next, while walking at a speed of 5 km/h, the cortical activation was measured without additional tasks for the first 2 minutes, immediately followed by the second dTMT session (session 4: on (4) of Fig. 1). The heart rate (HR) was recorded with a Polar H1 heart rate sensor (Polar Electro Europe AG, Switzerland) before, during, and after exercise. Participants’ fatigue immediately after the two types of exercise protocols was collected with the BORG scale (RPE, ratings of perceived exertion) [46]. Subsequently, participants completed the MPSTEFS questionnaire again. The third dTMT session assessed the sustained effects of each exercise protocol after a 10-minute break, but in a comfortable sitting position in this session (session 5: on (5) of Fig. 1).

In the present study, concentration changes in the amount of oxygenated and deoxygenated in the right and left: frontopolar area (FPA), dorsolateral prefrontal cortex (DLPFC), and motor cortex (M1) were recorded using a portable continuous optical fNIRS system (NIRSport 88, NIRx Medical Technologies LLC, New York, NY, USA). The setup consisted of 16 light sources and 16 detectors spaced approximately 3 cm between optodes to effectively compromise depth sensitivity and signal-to-noise ratio. The center of the NIRScaps for optode placement (EASYCAP GmbH, Herrsching, Germany) was placed over the vertex (Cz) according to the international 10–20 system by marking the midpoint between the nasion and the inion and the left and right preauricular points (see Fig. 3), to ensure experiment consistency of placement between participants and experimental sessions.

Fig. 3.

Configuration for fNIRS Measurement.

The data sampling rate of this fNIRS system is 7.81 Hz and includes two wavelengths of infrared light (760 and 850 nm) [56] to measure oxygenated and deoxygenated hemoglobin (oxyHb and deoxyHb) using the continuous-wave procedure (i.e., emitting infrared light from a source with a constant intensity and frequency) [57]. Data of oxyHb and deoxyHb were recorded on a tablet (Microsoft Surface Pro2 128 GB) with the NIRS Star 14 Software (NIRSport, NIRx Medical Technologies LLC, Glen Head, NY, USA). For each 30 seconds block of alternating task and rest, triggers were set using the NIRS Stim Software (NIRSport, NIRx Medical Technologies LLC, Glen Head, NY, USA).

In the present study, only the behavioral and neural data of dTMT-B condition were analyzed between the first and last session to investigate the extent to which two types of exercise affect the maintenance or improvement of cognitive performance at a behavioral and neuronal level after recovery on post-exercise (see Fig. 1). Besides, some data (e.g., demographic and physiological data, MPSTEFS, and behavioral and neural data on pre-and post-exercise) have already been reported in the work of [40]. However, behavioral and neural data were reprocessed and analyzed concerning the research question under investigation here.

Statistical analyses were performed using SPSS v. 27 (IBM Corp., Armonk, NY, USA). First, we explored the dependent variables for missing data points, normality of distributions (tested by Kolmogorov–Smirnov tests), and the presence of outliers. The effect sizes for all analyses are expressed using partial Eta${}^2$ ($\eta$${}_p$${}^2$) or Cohen’s d. For all statistical tests, an alpha level was set a priori p 0.05. The classification of the partial Eta${}^2$ also followed the conventions of Cohen [58]: 0.01 small effect; 0.06 medium effect; 0.14 strong effect. Sphericity issues were addressed with the Greenhouse Geisser correction for repeated measurements ANOVA. The statistically significant values in ANOVA lead to post-comparison using Bonferroni correction.

2.8.4 fNIRS Data

The analysis of the neural data was conducted with the NIRS Toolbox [59], an open-source program installed in Matlab 2021a (MathWorks, Natick, MA, USA). The relative coefficient of variation (CV in %) was first estimated with unfiltered raw data to check the data quality as a signal-to-noise ratio, a data processing method commonly used in NIRS measurements using multi-channel [60, 61]. CVs above 15% for each channel computed during each experimental session were excluded for the subsequent analysis [62]. After removing poor channels, data were processed using a low-pass filter with 0.2 Hz [57] to remove physiological components (e.g., heartbeats: 0.5 to 2.0 Hz, and respiration: 0.2 to 0.4 Hz). From these data, they could be converted to optical density. The converted data were then corrected for motion artifacts by attenuating outlier variations using temporal derivative distribution repair (TDDR) [63]. These preprocessed optical densities were finally converted into oxygenated hemoglobin (oxyHb) as well as deoxygenated hemoglobin (deoxyHb) concentration using the modified Beer-Lambert law [59]. The data of all channels were processed into time-series data by calculating the inter-trial mean of each dTMT condition. Each time-series data was corrected using the baseline period’s average value for two seconds before the onset of a task block (baseline correction). Furthermore, task contrast was performed by subtracting the time-series data for the resting conditions between task conditions from task conditions. Changes in hemodynamic responses for each condition were plotted as topographic on brain image over time during 30 seconds of task execution. Each channel of source-detector was averaged over six different regions of interest (ROI) in a further step for statistical analysis. The areas divided into six ROI consisted of the left frontopolar area (l-FPA) with source-detector pairs (S3-D2, S3-D5), the right frontopolar area (r-FPA) with source-detector pairs (S6-D5, S6-D7), the left dorsolateral prefrontal cortex (l-DLPFC) with source-detector pairs (S1-D1, S1-D2, S1-D3, S2-D2, S4-D3, S4-D4), and the right dorsolateral prefrontal cortex (r-DLPFC) with source-detector pairs (S5-D6, S7-D7, S8-D7, S8-D8), the left motor area (l-M1) containing all channels above the left motor cortex, and the right motor area (r-M1) containing all channels above the right motor cortex. We employed the freeware MNE-NIRS (v. 0.1.2; https://mne.tools/mne-nirs/) [64] from the MNE toolbox (v. 0.24.0; mne.tools) in Python [65] and MATLAB based toolbox NFRI (https://www.jichi.ac.jp/brainlab/tools.html) to visualize the topographic on a brain image [66]. It is believed that the oxyHb response can be used as a more sensitive indicator of changes in regional blood flow [67, 68]. For this reason, we used only changes of oxyHb levels for further statistical analysis. The use of ROIs as a factor in ANOVAs may lead to unintended statistical bias because the optical properties may differ systematically between ROIs [24]. Therefore, for statistical analysis, ANOVAs were performed separately for each ROI. To account for this, alpha values were corrected for the number of ROIs using the false discovery rate (FDR-corrected) to perform multiple comparisons [69]. The emphasis was also placed on the effect sizes when evaluating statistical significance due to the small number of participants in each group in the present study [70].

The parameters extracted from the fNIRS statistical analysis were the mean parameter and the slope over time (from onset to 32 sec) of the conditional block (see Fig. 4). The time of 32 sec was chosen to account for the fact that the maximum range of concentration changes is reached within a few seconds after the end of the task. According to the systematic review by Herold et al. [24], the methods used for statistical analysis of fNIRS data are very diverse (mean value, area under the curve, slope or GLM analysis, etc.). Therefore, for comparing the two parameters (slope and mean) in the present study, we followed the study of [39], which suggests that the slope method seems to be suitable for detecting changes in hemodynamic response over time with the cognitive workload.

Fig. 4.

Example for calculating slope for oxyHb while performing the dTMT over time.

In the statistical analysis, we analyzed 5 $\times$ 2 ANCOVAs with repeated measures for each oxyHb in ROI while adjusting for the rate of change in behavioral data (dTMT-B) as a covariate with the within-subjects factors “ session” (from first to last session) and the between-subjects factors “group” (HIIE and MCE), to test the effects of post-exercise recovery on changes of hemodynamic response over session. Also, we further calculated Pearson correlations between behavioral data and neural data in each session for each group separately.

Table 1 shows the demographic and physiological characteristics. No significant differences between the two exercise groups were found in the statistical analysis. Besides, the types of sports that subjects participate generally varied, including team sports (e.g., Basketball, Volleyball), endurance (e.g., rowing, track athletics, cycle, swim), individual sports (e.g., apparatus gymnastics, boxing, taekwondo), racket sports (e.g., tennis, table tennis).

Table 2 shows the physical/mental energy/fatigue components using a visual analog scale that yielded 12 separate ratings for Physical Energy, Physical Fatigue, Mental Energy, and Mental Fatigue before and after exercise. A 3 (“session”: before (1), before (3), and before (5)) $\times$ 2 (“energy or fatigue”: physical and mental energy/fatigue) $\times$ 2 (“group”: HIIE and MCE) showed a significant effect of session only for physical and mental fatigue, F (1.49, 40.2) = 3.38, p = 0.041, $\eta$${}_p$${}^2$ = 0.111, indicating an overall increasing physical and mental fatigue in post-exercise (before (5)) regardless of the group.

First, we performed 5 $\times$ 2 ANOVAs with repeated measurements on the number of connected circles, time spent inside circle, and speed between circles as a dependent dTMT-B variable with the within-subjects factors “session” from first to last session and with the between-subjects factor “group” (HIIE and MCE) to examine the change in cognitive performance over session. A significant effect of session was found for the variables “number of connected circles” F (2.92, 78.8) = 16.7, p 0.001, $\eta$${}_p$${}^2$ = 0.383, “time spent inside circle”, F (4, 108) = 8.78, p 0.001, $\eta$${}_p$${}^2$ = 0.245, and “speed spent between circles”, F (2.61, 70.4) = 8.57, p 0.001, $\eta$${}_p$${}^2$ = 0.241. However, no interaction effect between session and group was observed for any of these variables. A post hoc analysis of the factor session is shown in Fig. 5. No significant difference was found for the other variables, except for the speed between circles between the second session on the first day (during sitting) and the third session in pre-exercise (during walking). For the number of connected circles, we found a significant increase in cognitive performance between pre-and post-exercise, but there was no significant improvement between post-exercise (during walking) and -recovery (during sitting). When comparisons were made between the same conditions (during sitting), a significant difference was found between the second session on the first day and post-recovery.

Fig. 5.

Behavioral data extracted from the dTMT-B by group. * p 0.05.

In summary, for the behavioral data of the dTMT, task speed performance (speed between circles) increased during walking. Still, overall task performance (number of connected circles) remained unchanged between sitting and walking. Overall, cognitive performance improved significantly in both groups in post-exercise and was maintained in post-recovery.

Fig. 6 shows the hemodynamic responses during the performance of the dTMT for 30 seconds over a session. As shown in Fig. 6, for each condition (during sitting and walking), changes in oxyHb levels can be observed across sessions (from first to fifth session) during task performance. High cortical activation can be found after HIIE, and compared to MCE, high cortical activation was maintained to some extent after recovery.

Fig. 6.

Change of OxyHb for 30 seconds over session by group.

Table 3 shows the statistical results for the hemodynamic response over session with the ANCOVA with repeated measures for slope and mean of oxyHb in each ROI. The slope and mean value parameter over the entire session in all ROIs are shown in Fig. 7. The slope parameter showed a significant effect of the session in all ROIs, but the mean parameter showed a significant effect with a medium effect size in r-FPA, l-DLPFC, r-DLPFC, l-M1, and r-M1. An interaction effect with a medium effect size between session and group was observed for both parameters only in the l-M1 (see Fig. 8A), indicating a significant difference with an above medium effect size between groups (Fig. 7 (1-e and 2-e)) for slope value in post-exercise while walking (p = 0.073, d = 0.694) and post-recovery while sitting (p = 0.027, d = 0.871), and a significant difference between groups for mean value in post-exercise (p = 0.024, d = 0.886) and post-recovery (p = 0.026, d = 0.873). In summary, high hemodynamic responses were observed in all ROIs during performing the dTMT while walking in pre-and post-exercise and decreased again in post-recovery. Notably, a high hemodynamic response in l-M1 was maintained by HIIE in post-recovery compared with MCE.

Fig. 7.

Slope (1) and Mean (2) of OxyHb related to dTMT-B performance over session by group. Individual slope and mean oxyHb changes related to dTMT-B performance over the whole session are shown in all ROIs. Below each graph is the correlation coefficient between the rate of behavioral data (number of corrected circles) and the rate of neural data (slope and mean) between each session (e.g., $\mathrm{\Delta }$ session 2 – session 1, $\mathrm{\Delta }$ session 3 – session 2, $\mathrm{\Delta }$ session 4 – session 3, and $\mathrm{\Delta }$ session 5 – session 4).

Fig. 8.

Change of OxyHb related to dTMT-B performance for 30 seconds over session by group. (A) F-map of oxyHb signal changes of slope reflecting the interaction between session and group for the dTMT-B performance. Among the six regions of interest, an interaction with a medium effect size can be seen in the left motor cortex (l-M1) (FDR-corrected). F-value is displayed according to color bars. (B) Time course of oxyHb changes in response to dTMT-B performance in the l-M1 in post-exercise while walking between HIIE and MCE. (C) Time course of oxyHb changes in response to dTMT-B performance in the l-M1in post-recovery while sitting between HIIE and MCE. Error bars indicate standard error.

Demographic data were reported in our previous study [40]. Among them, there were no significant differences between groups. Also, no differences were found for VO${}_2$max at baseline, so that the group classification was already established before the intervention. As subjective variables, the fatigue of physical and mental parameters increased from pre- to post-exercise, but fatigue did not correlate with cognitive task performance and task-related cortical activation. Since these parameters were measured after the post-test and not immediately after exercise, they could not sensitively reflect physical and mental changes in energy/fatigue. However, as an objective variable, the change of heart rate induced by the exercise intervention during acute exercise and post-test led to a difference between groups [40]. This implies that the two different types of exercise-induced different intensities.

First, we conducted two sessions on the first day to effectively control for learning effects. No significant difference in cognitive performance for our three reported variables (see Fig. 5) was found between the first and second sessions. Recent studies have shown that individuals with high fitness levels show little or no improvement with acute exercise [71, 72]. However, due to the training regime we chose, the participants had to have at least an intermediate fitness level to perform the full extent of the training. In addition, it was possible to effectively control the learning effect by integrating a divide-and-combine approach [54], creating alternative sheets of the TMT task from the newly developed dTMT. Despite different conditions, there was no significant difference between the second and third sessions (between sitting and walking). In general, cognitive performance seems to be impaired during a dual-task while walking in children, adolescents, and older adults [73, 74]. However, this was not the case in our study. In comparing sitting and walking, the cognitive performance for the number of connected circles did not decrease due to the automation of gait in the treadmill task performance [75]. Notably, cognitive performance increased in the post-exercise period, which is supported by a study by Penati et al. [76] that showed better cognitive performance in fixed-speed walking on a treadmill than over-ground walking. Above all, the increased activation of the frontal lobe induced by exercise may have amplified the posture-second strategy [77] to focus more on cognitive tasks due to the automatized walking on the treadmill. As the posture-second strategy, dual-tasks while walking on a treadmill could enable resource allocation toward cognitive tasks while performing automatized walking naturally, contributing to improved motor and cognitive function. In this regard, dual-task performance on the treadmill may be a suitable alternative to improve motor and cognitive functions in clinical or rehabilitation fields.

However, significant cognitive performance increases were no longer observed post-recovery, but the performance was maintained. This sustained performance in the post-recovery period was significantly different, especially compared with the first and second session as baseline under the same condition (during sitting). According to the meta-analysis by Chang et al. [9], the greatest positive effect on cognitive performance was found after 10–20 minutes in post-exercise. This effect gradually decreased when it lasted longer than 20 minutes. In the present study, the increase and maintenance of cognitive performance in post-exercise and -recovery were independent of the intervention group. In the study by Tsukamoto et al. [13], the cognitive performance was also maintained for 20 minutes. In the case of HIIE, the performance was maintained up to 30 minutes longer than in MCE. However, in our study, only a 10-minute recovery was given, and thus a follow-up study on the post-exercise maintenance of cognitive performance should be conducted. Notwithstanding the limitations associated with the short recovery period, in the present study, in which exercise-induced changes of neural correlates related to task performance were additionally investigated using fNIRS, the comparisons between the two types of exercise should also be observed at the neural level.

To date, the effect of acute exercise on cognitive task-related neural correlates has been studied extensively using fNIRS (for a systematic review, see Herold et al. [24]). However, studies on the neural correlates of the maintenance of cognitive performance in post-exercise are still lacking. Furthermore, only few studies have investigated the sustained effects of exercise on task-related neural correlates [34, 35, 36, 37]. In general, acute exercise plays a role in increasing cortical activation concerning enhanced cognitive performance. Prefrontal cortical activation in post-exercise positively affects cognitive task performance [14]. In a recent study on the effects of exercise on cognitive performance in post-recovery, the cortical neural activation in the prefrontal area depends on age. At the behavioral level, inhibitory control for young and older adults improved immediately after 15-minutes of moderate exercise with 40–60% heart rate rebound, and, notably, was maintained even 30-minutes after recovery [34]. However, the cortical activation at the neural level appears to differ between young and older adults. In young adults, the cortical activation gradually decreased from baseline to post-exercise and -recovery periods. In contrast, in the elderly, it increased in the middle frontal lobe post-exercise and returned to pre-exercise levels in post-recovery again. In children aged 8–10 years, the cortical activation of the frontal lobe also increased in post-exercise (continuous and intermittent moderate-intensity exercise). However, it decreased steadily post-recovery, maintaining cognitive performance for 30 minutes [36].

The present study investigated the effects of exercise on various brain areas by additional measurement of the motor area, unlike other studies that are generally limited to frontal lobe measurements. Also, performing task conditions (during sitting and walking), type of cognitive task (dTMT-B), and exercise types (HIIE and MCE) should be considered when interpreting the results of the present study compared with other studies. Furthermore, from the point of view of statistical analysis, the simultaneous analysis and comparison of the mean and the slope value could be an appropriate alternative solution for the fNIRS analysis methods, for which there is no golden standard yet.

The performance of the 1st and 2nd sessions serves as a baseline to control the learning effect of the task. The results for oxyHb between the 1st and 2nd session showed decreased cortical activation only in the l-FPA with no change in cognitive performance. This appears to have been effectively utilized by reducing the neural resources due to the learning effect [78]. In comparing performance conditions, higher cortical activation in all ROIs was shown during the fine motor-cognitive task for 30 seconds while walking than sitting. In this regard, the higher cortical activation during walking compared with sitting might be due to allocating more neural resources to postural control and task performance while walking, which is why cognitive performance may have been maintained even during the dual-task condition. In particular, the slope of exercise-induced cortical activation was increased or maintained in post-exercise, which seems different between the two types of exercise. Notably, considering the mean value, there was a significant increase in the r-FPA and l-DLPFC from pre- to post-exercise, regardless of the intervention group, which is consistent with other studies showing exercise-induced increased cortical activation in the l-DLPFC [27, 79]. However, compared with these studies, more neural resources may need to be recruited in other areas due to type of exercise (intensity or duration), performing condition (sitting or walking), and type of cognitive task (Stroop task or fine-motor cognitive task) to maintain task performance in post-exercise. Thus, further studies should be conducted considering the factors mentioned above.

More importantly, depending on the characteristics of the fine-motor cognitive task, the motor area may have been significantly differently affected by the two types of exercise. After HIIE, the cortical activation was increased in both M1 areas, whereas it decreased after MCE. Notably, the rate of change for oxyHb (both parameters for slope and mean) between pre-and post-exercise (adjusting rate of change for dTMT performance as a covariate) showed a significant difference between groups. The increased activation of M1 in the HIIE group, along with an increase in the frontal lobe, could be explained by the need for more neural resources for postural control and fine-motor control, along with smooth task performance while walking on the treadmill [40]. According to the conceptual model of mechanisms of physical activity [1, 80], regular exercise induces changes in cellular and molecules such as brain growth factors, which in turn lead to structural and functional changes in the brain, leading to an increase in cognitive performance. In addition, the upregulation of VEGF by exercise in animal models increased cerebral blood flow [81, 82]. Moreover, the increased cerebral blood flow velocity is related to fitness level and healthy human aging [83]. Given the physical fitness level of the subjects participating in the regular exercise in the present study, the increase in cerebral blood flow to the frontal lobe in post-exercise, which plays an essential role in the executive function, seems reasonable. A further increase in M1 caused by HIIE could also be explained. Therefore, other studies are needed to determine whether HIIE induces high cortical activation and enhanced cognitive performance when used in different age groups (children and the elderly) or groups with varying levels of fitness (high and low) [71].

Moreover, cortical activation had returned to baseline levels (1st and 2nd sessions) in all ROIs after recovery, even when both variables for the slope and mean parameters were accounted for. Notably, interaction with the above medium effect size for both slope and mean parameters was found only for l-M1, indicating that the higher cortical activation in the l-M1 appeared to be maintained in HIIE than MCE. In addition, in oxyHb, there was no difference between groups in the rate of change between post-exercise and -recovery periods. However, the exercise-induced increased cortical activation from post-exercise was maintained until the post-recovery period, resulting in a significant difference between groups (p = 0.027, d = 0.871). In a previous study that applied moderate-intensity exercise [84], the cortical activation of the frontal lobe tended to decrease in post-recovery. Still, the cognitive performance was maintained for up to 30 minutes. In the present study, decreased cortical activation was also observed in the frontal lobe. An additional measurement of the motor cortex also revealed decreased oxyHb in the left and right M1. In particular, the left M1 was more affected by HIIE. This can be interpreted as the effect of exercise intensity. Although it was necessary to mobilize a large amount of neural resources to process the fine-motor cognitive task simultaneously with walking in post-exercise, a general downregulation of cortical activation occurred during sitting in the post-recovery period. HIIE, which may have led to fatigue due to high intensity, could affect the upregulation of the left M1 in processing the fine motor task.

Also, the change in behavioral data was already considered by performing ANCOVA with repeated measurements between sessions and groups for oxyHb, adjusting the $\mathrm{\Delta }$ of the first and last sessions for dTMT performance. As for any correlation between the $\mathrm{\Delta }$ in hemodynamic responses and the $\mathrm{\Delta }$ in dTMT performance (number of correct connected circles) (see Fig. 7 under each graph), except for strong correlations with behavioral data and hemodynamic response (mean parameter) of the left frontal lobe for HIIE group between pre-and post-exercise (r = 0.698, p = 0.006), no significant correlation was found between sessions. HIIE exercise resulted in an increase in the left frontal lobe and improved cognitive performance. In contrast, in the MCE group, behavioral data and cortical activation increased simultaneously, although there was no significant correlation. Overall, the absence of significant correlations could result from the proper allocation and effective use of neural resources depending on the execution and task conditions in the experimental design. In addition, the $\mathrm{\Delta }$ of cortical activation decreased without any difference between groups in all ROIs from post-exercise to -recovery. In contrast, the $\mathrm{\Delta }$ of the behavioral data resulted in a difference with small to medium effect size (p = 0.202, d = 0.486) between HIIE (12.3 $\pm$ 27.2%) and MCE (1.6 $\pm$ 24.2%). For this reason, in the present study involving healthy young adults with a high fitness level according to normative data for VO${}_2$max (at least a fair or good fitness for VO${}_2$max from Heywood [85]), there was an increase in cognitive performance from pre-exercise to post-exercise and maintenance occurred in post-recovery, resulting in different cortical activations adapted to the two types of exercise.

Although the present study has provided novel attempts and interesting findings, some limitations are worth noting. First, the sample size in our study was small, which may indicate high inter-individual variability in exercise-induced cortical activation. Therefore, in statistical analysis, emphasis was placed on the effect size to compensate for this. Nevertheless, a large number of participants is required for generalization. Second, although various moderator parameters between acute exercise and cognition are involved, it is necessary to observe the acute and sustained effects of cortical activation induced by exercise on cognitive function through studies that consider various fitness levels (age group or high/low fitness). For this reason, given the participants’ high fitness levels, we probably did not observe any differences in cognitive performance at the behavioral level due to the effective use of neural resources by the two types of exercise. Finally, our study’s recovery duration was very short compared to studies investigating the sustained effect of acute exercise-induced neural correlates on cognitive function. Since the regular accumulation of acute exercise results from chronic exercise, the sustained effect of acute exercise by prolonging the recovery duration could be a bridge to show the effect of chronic exercise.