The study involved 36 healthy young college student volunteers (18 women; aged 21 ± 1 year) (Table 2). Out of 37 initial participants, one was excluded due to medication affecting the central nervous system. The sample size, set at 12 per group, was determined by a power analysis referencing Tanaka et al. (2009), and aligns with the recommended minimum for pilot studies (Julious, 2005).

Participants had no history of orthopedic or neurological diseases and were not treated with medications that would affect the central nervous system. To control factors that could influence the effectiveness of non-invasive brain stimulation, participants were asked to avoid vigorous physical activity and consumption of alcohol and caffeine during the experiment period and to ensure adequate sleep the night before the experiment to avoid sleep deprivation (Guerra et al., 2020).

All participants gave written informed consent before participating in the experiment. This study was approved by the Ethics Committee of Yamagata Prefectural University of Health Sciences (approval number: 1806–06) and was performed according to the ethical standards of the Declaration of Helsinki.

The study was a randomized, triple-blind, sham-controlled trial. Intervention conditions were concealed from participants, outcome assessors, and data analysts. Participants were stratified by sex and randomly assigned to groups using float numbers between 0 and 1 from a continuous uniform distribution. Assignment was based on whether the number drawn is smaller or equal in value compared to 0.5 or larger than 0.5. These numbers were generated by a third party unrelated to evaluation or intervention (Microsoft, Washington, United States).

To induce the maximal EFs in both SMA and M1, the optimal electrode configuration was determined through computer simulations (see “Electric field simulation”). These simulations indicated that the field maximizing electrode-pair positions were at FCz and POz (Figure 1). Participants received tDCS with one of the following three different electrode placements: (1) anodal and cathodal electrodes positioned over FCz and POz, respectively, resulting in anterior-to-posterior current flow (A-P tDCS); (2) the reverse arrangement, where anodal and cathodal electrodes were positioned over POz and FCz, respectively, resulting in the reverse (posterior to anterior) currents (P-A tDCS), and (3) no current passed (sham tDCS) (Figure 2).

Participants were stratified by sex and were randomly assigned to one of the above three groups. All participants underwent the sit-to-stand training following tDCS intervention. The sit-to-stand training was conducted once every 3 or 4 days for 3 weeks (a total of 5 sessions). Exercise performance, measured by the number of sit-to-stands executed in a minute, was assessed pre-training, during each training session, and post-training. Muscle strength (knee extension and flexion peak torques) was measured pre- and post-training. Cortical excitability was measured pre-training, immediately after the first training session, and post-training. Pre-training assessments were conducted between 48 and 72 h before the first training session. Post-training assessments were conducted between 48 and 72 h after the final training session.

tDCS was administered using a DC-Stimulator-Plus (neuroConn, GmbH, Germany) connected to a pair of sponge surface electrodes, each with a surface area of 35 cm², soaked in a 0.9% NaCl saline solution. With the electrodes placed on the scalp of the participant, a direct current of 2 mA was applied for 10 min. Skin pre-treatment agents and alcohol swabs were used to reduce scalp skin resistance at the electrode contact area. The electrode placement was consistently maintained at the same positions, determined by measuring the length of the participant’s head with a tape measure during each session. For the sham condition, the same procedure was performed; however, the current was turned off after the first 15 s to mimic the transient skin sensation felt at the beginning of the direct current. Intervention condition was masked to participants, outcome assessors, and data analysts.

In order to induce the maximal EF strength in both SMA and M1, the optimal electrode montage was determined based on computer simulation of the EFs, similarly to our previous studies (Laakso et al., 2016; Fujimoto et al., 2017).

For the computer simulations of the EF, we utilized 62 individual MRI-based anatomical models, consistent with those used in our previous studies (Laakso et al., 2016; Fujimoto et al., 2017). Each model, with an isotropic resolution of 0.5 mm, was constructed from the segmentation of T1- and T2-weighted (T1w and T2w) MRI data. FreeSurfer image analysis software was employed to segment the brain, including gray and white matter (Dale et al., 1999; Fischl et al., 1999; Fischl and Dale, 2000). Other tissue compartments were segmented as follows: The inner and outer boundaries of the skull, along with the outer surface of the scalp, were identified from the MRI data. An experienced investigator utilized semi-automatic image processing techniques, including thresholding, opening/closing, smoothing, and region growing on the T1w and T2w data, with the parameter values chosen on a per-subject basis. The spaces between the brain, inner skull surface, outer skull surface, and the scalp surface were automatically segmented into multiple tissue types using both MRI data and geometrical operations. The space between the skull and brain was segmented into CSF (large T2w), blood (small T2w), and dura (outer 1.5 mm which is not CSF or a large dural vein). The space between the inner and outer boundaries of the skull was segmented into compact bone (small T2w but having at least a 1-mm-thick inner layer and 1.5-mm-thick outer layer) and spongy bone (large T2w). Lastly, the space between the outer boundary of the skull and the surface of the scalp was stratified into fat (large T1w), muscle (small T1w), and skin (small T1w, at least 2-mm thick, at most 1-cm thick including subcutaneous fat) The quality of each model was verified by two independent examiners (Laakso et al., 2016). The electrical conductivities assigned to each tissue were (Laakso et al., 2016): gray matter (0.2 S/m), white matter (0.14 S/m), blood (0.7 S/m), compact bone (0.008 S/m), spongy bone (0.027 S/m), dura (0.16 S/m), CSF (1.8 S/m), muscle (0.16 S/m), skin and fat (0.08 S/m), and eye (1.5 S/m). The sensitivity of the EF to conductivity values was assessed by varying bone conductivity by ±50% and the CSF conductivity by ±10% and repeating all calculations.

The EF was determined as the gradient of the electric scalar potential ( ϕ ) , which was obtained by numerically solving ∇ . σ ∇ ϕ = i with homogeneous Neumann boundary conditions, where σ is the conductivity (S/m) and i is the current source/sink (A/m³). The finite element method (FEM) with 0.5 mm cubical first-order elements was employed for numerical solution (Laakso and Hirata, 2012; available at https://version.aalto.fi/gitlab/ilaakso/vgm-fem). The stimulation electrodes were modeled identically to our previous studies (Laakso et al., 2016; Fujimoto et al., 2017), as a 1 mm thick rubber sheet embedded in a saline-soaked sponge (1.6 S/m), with size, shape, and current intensity identical to those in the actual experiment (Figure 1). After obtaining the EF, it was interpolated to a polygonal surface reconstruction of the brain at 1 mm depth below the gray matter surface. As detailed in Laakso et al. (2016), the individual surface EFs were then registered with each other and mapped to the Montreal Neurological Institute (MNI) ICBM 2009a nonlinear asymmetric template (Fonov et al., 2009, 2011) using FreeSurfer and the spherical demons algorithm (Yeo et al., 2010). This process enabled determination of the population-average EFs of the bilateral SMA and M1 in the MNI space.

For the optimization of the electrode montage, we considered four regions of interest. The regions of interest for M1 and SMA had a radius of 1 cm and were centered at [± 9, − 39, 54] and [± 3, − 9, 60] in the standard brain space, respectively. Our aim was to find a montage of two electrodes that produces symmetric bilateral stimulation and maximizes the average EF strength over the four regions of interest. To achieve symmetric stimulation, both the anode and cathode needed to be placed on the midline of the head. For practical applicability, the electrode locations were selected using the International 10–10 system: The first electrode was placed at Fz, FCz, Cz, CPz, or Pz, and the second electrode at an extracephalic location, frontally (Fpz), posteriorly (Iz), or close to the stimulated areas (POz or Pz). The electrodes were oriented so that the long edges of the electrodes were perpendicular to the posterior–anterior direction. Any locations which would have caused the electrodes to overlap were excluded, leaving 23 electrode montages, listed in the first two columns of Table 1.

The EF produced by each montage were calculated for all 62 head models, and the EF strength averaged over each/all regions of interest was determined. Finally, the mean and standard deviation of these average values were calculated over the 62 head models. The results for each studied electrode montage are provided in Table 1. It was found that the FCz-POz configuration induced the largest mean value of the average EF strength over the four regions of interests, closely followed by the Fz-Pz montage. At the individual level, FCz-POz was the optimal montage in 32/62 models and Fz-Pz in 22/62 models (Supplementary Table S1). Repeating the EF analyses for altered bone and CSF conductivities indicated that either the FCz-POz or the Fz-Pz montage was the optimal configuration regardless of the choice of the conductivity values (Supplementary Table S1). Thus, the FCz-POz configuration, visualized in Figure 1, was used in the experiment.

The participants performed the sit-to-stand task starting from a seated position on a 20 cm box. They were instructed to perform the sit-to-stand task by standing up straight from the box as quickly as possible with fully extended trunk, hip joints, and knee joints, then sit down. The training protocol of a single session consisted of 5 sets of 1 min sit-to-stand task with a 180-s rest period between sets (Figure 2). While monitoring the participant’s performance to ensure proper form (fully extend the trunk, hip joints, and knee joints) examiners counted the completion of each correctly executed sit-to-stand. The counts were averaged over the 5 sets. The sit-to-stand training was conducted a total of 5 sessions. The 48 to 72 h interval between sessions was used to avoid the effect of fatigue on the performance. Before training session, participants warmed up on a bicycle for 5 min, then performed a set of stretches focused on the knee extensor and flexor muscles. They repeated the same routine to warm down after each session.

Knee extensor and flexor peak torques were assessed using an isokinetic dynamometer (Multi-Joint 3, Biodex Medical Systems Inc., Shirley, NY, United States) with the same procedure as Maeda et al. (2017). Participants were instructed to extend the knee with maximal effort while the dynamometer flexed the knee at a speed of 30°/s from the initial 20° to the 90° eccentric contraction of the knee extensors. For the measurement of the knee flexors torques, participants flexed the knee while the dynamometer extended the knee at 30°/s from 90° to 20°. The maximal knee extensor and flexor torques were evaluated for 3 sets (5 repetitions/set) under eccentric (30°/s) conditions. The maximal knee extensor and flexor torques obtained in 5 repetitions were taken as the peak torque, and the average of the peak torque for each set was calculated.

The participants were comfortably seated in a chair with their arms resting on a cushion. The electromyography (EMG) was recorded via Ag/AgCl-plated surface electrodes (1 cm diameter) placed 2 cm apart over the right rectus femoris (RF) muscle. Responses were acquired using a Neuropack MEB-2200 system (Nihon Kohden, Tokyo, JPN) filtered in the 10 Hz to 1 kHz pass-band. EMG signals were sampled at 5 kHz and stored on the computer for off-line analysis using the LabVIEW software (National Instruments Inc., Austin, Texas, United States).

To assess changes in motor cortex excitability, single-pulse transcranial magnetic stimulation (TMS) was applied to the leg area of the left M1 using a magnetic stimulator (Magstim200, Magstim, Dyfed, UK) connected to a double-cone coil of 110-mm diameter. The hotspot of the M1 was confirmed based on induction of the largest MEP amplitude in the right RF muscle during tonic voluntary contraction. The stimulation intensity was adjusted to 120% of the active motor threshold (aMT). The aMT was defined as the lowest stimulus intensity needed to produce MEPs greater than 200 μV in at least 5 out of 10 consecutive trials during the maintenance of 100 μV of RF voluntary isometric contraction (Rossini et al., 2015; Temesi et al., 2017). The time between stimulus pulses was varied between 5 and 7 s. The stimulus timing was automatically controlled using LabVIEW.

In order to induce SICI we used a subthreshold conditioning paired-pulse paradigm (Kujirai et al., 1993). We used stimulus intensities of 80% aMT for the conditioning stimulus and 120% resting motor threshold (rMT) for the test stimulus. Throughout the experiment, the test stimulus was adjusted to maintain the MEP amplitude equal to the RF MEP amplitude at baseline. The interstimulus interval was set at 2.5 ms, and 15 MEPs were recorded from the RF muscle (Fisher et al., 2002). The conditioned MEP amplitudes were expressed as percentages of the mean test MEP amplitudes.

The primary outcome measures included the sit-to-stand counts and muscle strength measured pre-training and post-training. We used the 2-way mixed-model analysis of variance (ANOVA) to evaluate the differences in outcome with group (A-P tDCS, P-A tDCS, sham tDCS) and time (pre-training and post-training) used as within-subject factors. Muscle strength was tested separately for the left and right knee flexors and extensors. Similarly, for the sit-to-stand counts during training sessions, we used a 2-way mixed-model ANOVA with the group (A-P tDCS, P-A tDCS, sham tDCS) and time (pre-training, session 1, session 2, session 3, session 4, and session 5) as factors. A t-test with Bonferroni adjustment for multiple comparisons was performed to compare training effects for the group and time factor. For MEP amplitudes and SICI, we applied a 2-way mixed-model ANOVA with the group (A-P tDCS, P-A tDCS, sham tDCS) and time (pre-training, immediately after a first session of training, post-training) as factors. A t-test with Bonferroni adjustment for multiple comparisons was performed to compare changes in cortical excitability pre-training, immediately after the first session of training, post-training for the group and time factor. p values <0.05 were considered statistically significant for all analyses. Statistical analyses were performed using SPSS 28.0 (IBM, Armonk, NY, United States).

The sit-to-stand counts continued to increase throughout the 5 sessions and were greater after than before training in all groups. At post-training assessment, the sit-to-stand counts for the A-P tDCS group were significantly higher than the sham tDCS group, indicating that A-P tDCS promoted sit-to-stand performance (Figure 3). These observations were supported by primary outcome results of the 2-way mixed-model ANOVA, which revealed significant interactions (F_{(2, 33)} = 3.652, p < 0.05), with a significant main effect of time (F_{(1, 33)} = 88.990, p < 0.01). No significant main effect of the intervention was observed (F_{(2, 33)} = 0.398, p = 0.675). Similarly, the multiple comparison test revealed that the sit-to-stand counts were significantly increased at post-training for all groups compared to the pre-training (p < 0.01). When the sit-to-stand counts were compared between the groups, the counts were significantly greater in the A-P tDCS group compared to the sham tDCS group (p < 0.05) at post-training. However, 2-way mixed-model ANOVA found no significant differences in the counts among the groups during training session (F_{(10, 165)} = 1.879, p = 0.051). The main effects of time were significant (F_{(5, 165)} = 25.685, p < 0.01), but no significant main effect of the intervention was observed (F_{(2, 33)} = 0.777, p = 0.468).

The peak torque in right and left knee flexors were increased in the P-A tDCS group, and only in that group, following 3 weeks of training (Figure 4). In contrast, no changes were observed for peak torque in the knee extensors in any of the groups. Statistical analyses supported these observations. For knee flexor torques, 2-way mixed-model ANOVA showed significant interactions between intervention and time (Right: F_{(2, 33)} = 8.099, p < 0.01, Left: F_{(2, 33)} = 10.917, p < 0.01). The main effects of time were significant (Right: F_{(1, 33)} = 9.425, p < 0.01, Left: F_{(1, 33)} = 8.436, p < 0.01). No significant main effects of the intervention were observed (p > 0.1 for both right and left). For knee extensor torques, there was also no interaction between intervention and time, and the main effects of both time and intervention were not significant (p > 0.1 for each test, for both right and left).

Multiple comparison test revealed that the left and right knee flexor peak torques in the P-A tDCS group were significantly increased at post-training above their pre-training levels (p < 0.01). The left knee flexor peak torques in the P-A tDCS group, in particular, were significantly increased over levels in the sham tDCS group at post-training (p < 0.05).

There was no significant interaction between intervention and time (F_{(4, 66)} = 0.655, p = 0.626), and the main effects of both time and intervention were not significant (p > 0.3 for both). These results confirmed that MEP amplitudes remained stable at pre-training levels during and after training.