We simulated 300 brain MRIs derived from 3 age groups (young, middle, and older) across 2 sexes (male and female) for the two montage positions (F3-RSO and CP5-Cz) and two current doses (1 and 2 mA). For each simulated head model, the DTDI was calculated as the measure of focality and thresholded at 0.75 to segregate individuals into two groups (focal and non-focal). The volume of CSF, WM, and GM from 116 brain regions was obtained for each MRI. Finally, the association of focality with regional volume and its dependence on age and sex were evaluated. A similar analysis was performed on the scores of 4 behavioral measures (cognition, anxiety, depression, and sleep) to investigate the phenotypic signature of focality.

The dataset used in the study was taken from publicly available Cambridge Centre for Aging and Neuroscience (Cam-CAN) inventory (Shafto et al., 2014; Taylor et al., 2017).¹ The Cam-CAN project uses epidemiological, cognitive, and neuroimaging data to understand the aging brain. The repository provides a subset of data from 700 English-speaking healthy individuals (age range 18–88 years) with mini-mental state examination (MMSE) score above 27, having normal hearing and vision, and free from neurological or psychiatric conditions. We selected the T1-weighted brain MRI of 300 right-handed individuals (150 male) such that the images can be divided into three (uniformly spaced) age groups with 100 age-sex matched individuals (50 males and females) in each group comprising of—(a) Young adults (20 ≤ × < 40 years), (b) Middle adults (40 ≤ × < 60 years), and (c) Older adults (60 ≤ × < 80 years). The age range selected in the study was adopted from our previous work (Kashyap et al., 2021) as we observed focality to change significantly in males from middle age onwards. Consequently, in this work, we intended to explore the volumetric changes in the brain anatomy that are innate to each group and associate with the focality-based changes. The MRIs were collected from a 3T Siemens TIM Trio scanner with a 32-channel head coil using MPRAGE sequence (TR = 2,250 ms, TE = 2.99 ms, flip angle = 9^°, Voxel size = 1 × 1 × 1 mm³, FOV = 256 × 240 × 192 mm³, GRAPPA: 2, TI: 900 ms).

The tDCS electrode placement followed the 10–20 electroencephalogram (EEG) labeling system to convey the position of anode and cathode over the scalp. Two montages with electrodes (Anode-Cathode) positioned at F3-RSO and CP5-Cz were simulated using ROAST (version 3.0) toolbox (Huang et al., 2019). For each MRI, the montages were simulated for two current doses (1 and 2 mA). Altogether, four tDCS configurations—(1) F3-RSO for 1 mA, (2) F3-RSO for 2 mA, (3) CP5-Cz for 1 mA, and (4) CP5-Cz for 2 mA were simulated across 3 age-groups with equal males and females. For all configurations, the electrode sizes were kept constant at 5 × 5 cm² and the conductivity values for the various brain tissues were set at default [as mentioned in ROAST (Huang et al., 2019)] with WM (default 0.126 S/m); GM (default 0.276 S/m); CSF (default 1.65 S/m); bone (default 0.01 S/m); skin (default 0.465 S/m); air (default 2.5×10⁻¹⁴ S/m); gel (default 0.3 S/m); electrode (default 5.9×10⁷ S/m) (Burger and van Milaan, 1943; Geddes and Baker, 1967; Baumann et al., 1997; Haueisen et al., 1997; Akhtari et al., 2002; Nadeem et al., 2003). The ROAST segments the brain MRI and provides the simulated current density distribution for the virtual tDCS electrodes placed over the scalp. The simulated distribution was found to correlate with in-vivo electrophysiological measurements (Huang et al., 2017). The ROAST outputs were the location [x, y, and z coordinates (in mm)] of each brain area in the native space and the magnitude of current density (mA/m²) values corresponding to the location. In total, four tDCS configurations (2 montage positions × 2 current doses) applied on 300 individuals produced 1,200 simulations (300 individuals × 4 tDCS configurations) with target ROI at DLPFC for montage F3-RSO, and at IPL for montage CP5-Cz.

Although there are many possible combinations of tDCS montages and currents that could have been used for the simulations in the present investigation, the primary reasons behind our choice for the 4 tDCS configurations (2 montage positions and 2 current doses) that we used are- (1) Encompassment of inter- and intra-hemispheric configurations- In tDCS, the current travels between anode and cathode via the conductive brain tissue (Burger and van Milaan, 1943; Geddes and Baker, 1967; Akhtari et al., 2002; Nadeem et al., 2003). For the montage F3-RSO, the electrodes are placed across the two hemispheres of the brain. In this case, the current must percolate through the subcortical structures to reach the opposite hemisphere (Miranda et al., 2006; Bikson et al., 2010; Moliadze et al., 2010). In contrast, the montage CP5-Cz is on the same hemisphere, and the current most likely flow superficially (i.e., via the surface of the cortex). We anticipated that the different flow of current in the two montages will manifest discrete electric field patterns. Thus, the two montages can provide a representative understanding of the influence of regional volumetric parameters on focality of tDCS. (2) Utility of the two montage positions- Numerous studies have recommended the montage F3-RSO for the treatment of depression (Ammann et al., 2017; Moffa et al., 2020; Razza et al., 2020) and working memory (Nikolin et al., 2018). Similarly, CP5-Cz montage that stimulated the left IPL was used to modulate the reading behavior (Bhattacharjee et al., 2019a,b, 2020, 2022). Therefore, exploring focality with the two montage positions can have clinical importance. (3) Benefits of the two current doses- tDCS is considered safe for current doses up to 2 mA, and the vast majority of studies use intensity of either 1 or 2 mA (Jamil et al., 2017; Reinhart et al., 2017; Thair et al., 2017). Hence, the montages simulated for two widely practiced current doses will have broad applicability.

The i-SATA converts the coordinates of the brain from native space (obtained as the output of ROAST) to the standard space. There are two versions of i-SATA- (1) i-SATA (Talairach) (Kashyap et al., 2020) that converts to the Talairach brain space using the Talairach client (Lancaster et al., 2000), and (2) i-SATA (MNI) (Kashyap et al., 2021) that converts to Montreal neurological institute (MNI) space (Collins et al., 1994; Amunts and Zilles, 2001) using the 116 region-based AAL atlas from SPM anatomy toolbox (Eickhoff et al., 2005). In this work, i-SATA (MNI) was utilized to demarcate the anatomical boundaries of the 116 ROIs and estimate their current density. The magnitudes of current density received by the coordinates of an ROI were averaged to represent the current density of the region. DTDI is the ratio of the current density value at the target ROI to the peak current density value formed at any intermediary region.

The DTDI ranges from 0 (when no current reaches the Target ROI) to 1 (when the peak is formed at the target ROI), and values above 0.75 ensure that a considerable amount of current is reaching the target ROI.

For the montage F3-RSO with two current intensities (1 and 2 mA), the DTDI and the current density at the left DLPFC (target ROI) were obtained. Similarly, for the montage CP5-Cz and the two current doses, DTDI and current density were calculated for left IPL. For any tDCS configuration, two groups were delineated based on the DTDI value- (1) Focal group- which constituted individuals whose DTDI ≥ 0.75, and (2) Non-focal group- which represented individuals with DTDI < 0.75.

The 300 MRIs were preprocessed in CAT (version 12.8) using the pipeline (with default parameters) inbuilt for the region-based morphometric analysis (Gaser and Dahnke, 2016). In this pipeline, the regional tissue volumes for an ROI defined in an atlas brain were obtained after mapping it to the individual brain using high dimensional spatial registration. The images were corrected for bias–field inhomogeneity, spatially normalized, and smoothened by an isotropic Gaussian kernel of 6 mm full width at half maximum. The partial volume estimation (Tohka et al., 2004) technique was used to segment the MRI computed the regional volume. This method considered every voxel in the image to be a mixture of three tissues, namely the GM, WM, and CSF. The fraction (p1, p2, and p3) of GM, WM, and CSF in each voxel was calculated such that p1 + p2 + p3 = 1. This was followed by the volumetric estimation of the ROIs in the native space. The 116 region-based AAL atlas (Tzourio-Mazoyer et al., 2002) was then warped from the standard MNI space to the native space using the inverse deformation field derived for each individual. The regional volume of GM, WM and CSF was then estimated after the fractional value of each voxel was multiplied by the voxel size. For example, the GM volume for an AAL demarcated ROI is the sum of p1 values contained within that ROI multiplied by the voxel size (similarly for WM and CSF volumes). The CSF-filled region can have a small amount of WM or GM volume (and vice-versa) owing to the partial volume effect. Typically, the GM/WM volumes for CSF regions are quite small compared to their corresponding GM/WM volumes. For instance, the CSF-filled ventricles can have very small values near the edges that account for the GM volume of that ROI. The regional volumes are in milliliters (ml) but were scaled by the total intracranial volume (Total volume of WM + GM + CSF) to correct for the variation in the brain size (Cousijn et al., 2012). Initially, age and sex were not regressed out since the association of these two factors with the volumetric parameters were of interest in the ANOVA analysis of the two groups (focal and non-focal, for details refer to the section- Statistical Analysis of variation of focality). However, studies have reported that age and sex can influence the volumetric results (Cousijn et al., 2012). Therefore, ANOVA analysis was repeated after regressing out age and sex from the volumetric parameters.

Much research has been devoted to the exploration of behaviors that are modulated by tDCS. A majority of the studies have found convincing evidence of the ability of tDCS to modulate cognition, anxiety, depression, and sleep (Ammann et al., 2017; Thair et al., 2017; Dondé et al., 2018; Moffa et al., 2020; Razza et al., 2020; Yamada and Sumiyoshi, 2021). Interestingly, the Cam-CAN repository provides scores for these behavioral measures- (1) MMSE for cognition, (2) hospital anxiety and depression scale (HADS) for anxiety (HADS_anxiety), (3) HADS for depression (HADS_depression), and (4) sleep quality (Pittsburg sleep quality index, PSQI) for the 300 individuals considered in the study. This motivated us to explore whether any of these behavioral measures exhibited significant differences between the focal and non-focal groups. Investigating the phenotypic trend expressed in focality can be useful when an MRI of a subject is not available (or cannot be done), and there is an intention to identify if the subject would receive focal stimulation for the chosen tDCS configuration.

For each tDCS configuration, the individuals from three age groups (Young, Middle, and Older) across both sexes (males and females) were segregated into the focal and non-focal groups. For each individual, the regional CSF, GM and WM volumes were extracted from 116 ROIs. For each montage (2 of them), for each tDCS current value (2 of them), and for each ROI (116 of them), we performed three 3-way ANOVAs separately, one with GM as the dependent variable (DV), one with WM as the DV, and one with CSF as the DV. Each of the ANOVAs had between-subject factors such as age group, sex, and focality group as independent variables. For a given ROI, if any of the 3 DV’s showed a significant (p < 0.05, Bonferroni corrected) main effect or interaction involving the focality group, then that ROI was to be considered as a “focal ROI.” For each group, the Shapiro-Wilk test was performed to verify the assumptions of normality, and for each ANOVA, Mauchly’s test evaluated the assumption of sphericity. Post hoc analysis was performed to further characterize the main and interaction effects of focality. Finally, the current densities at the target ROI were correlated (using Pearson correlation) with the volumetric values of focal ROI. To check whether the correlation has an issue of heteroscedasticity, we used the White test. Additionally, for each tDCS configuration (montage and current dose), we also correlated the CSF volume of the focal ROI with the current density values of the target ROI for the focal and the non-focal group separately. If the correlation coefficient is significantly (p < 0.05) different between the two groups, then it will indicate that the focality of stimulation is influenced by the regional anatomy of the brain. Finally, we performed ANOVA analysis on the four behavioral measures (MMSE, HADS_anxiety, HADS_depression, PSQI).

The DTDI quantified focality in tDCS and was well integrated with the i-SATA (MNI) toolbox (Kashyap et al., 2021). This metric was inspired by a similar measure widely followed in nerve stimulation techniques where the targeted region is considered to be adequately stimulated for values above the threshold of 0.75 (Deurloo et al., 2000; Tigra et al., 2016; Kosta et al., 2019). Since such a high threshold ensures that the target ROI will be effectively stimulated, we followed the same value (0.75) as the threshold of DTDI to segregate the focal and non-focal groups. We understand that the threshold plays a vital role, and there is a need to identify the limit until which our findings may remain valid. To examine this, we took an iterative approach wherein we reduced the threshold of DTDI in steps of 0.01 (from the previous threshold) and repeated the 3-way ANOVA analysis (for the volume parameter). The threshold at which null/insignificant findings appeared, the iteration stops. For example, in the first iteration, we separated the focal and non-focal groups for a threshold of 0.74 (i.e., 0.75–0.01) and performed the ANOVA analysis. We checked the consistency of the results of ANOVA obtained for the DTDI threshold of 0.74 with the findings obtained previously with a threshold of 0.75. If the results were consistent (i.e., p-values are significant), then in the second iteration, we repeated the previous steps for the threshold of 0.73 (i.e., 0.74–0.01). Otherwise, we stopped iterating further.

The focus of the study was to investigate the contribution of regional volume to differences in focality across the two montages and two current intensities. Interestingly, the results were driven by montage position and current intensities do not alter them. For the F3-RSO configurations and two current intensities, the CSF volume at the left caudate nucleus was significantly higher in the focal group (for the older group only). The CSF volume correlated positively with the current density at the left DLPFC. For the configurations with electrode position at CP5-Cz and two current intensities, a higher CSF volume was observed for the non-focal group at the left pre- and post-central gyrus. The correlation between the CSF volume and the current density at the left IPL was negative. Surprisingly, the results from the two montages (F3-RSO and CP5-Cz) appeared to contradict each other.

The key motive of the study was to comprehend focality and its interaction with age and sex. To this, the current densities were simulated for all individuals. As an example, we show (Figure 1A) the spread of the simulated current density over an individual’s brain along with the montage electrodes placed at F3 and RSO. The DTDI of all individuals were estimated for the target area at left DLPFC. The individuals were divided into focal and non-focal groups based on the DTDI threshold prefixed at 0.75. Table 1 shows the range (Mean ± Std) of DTDI for the two groups (focal and non-focal) across the two tDCS configurations (1 and 2 mA). The number of males and females in each group (focal/non-focal) and their distribution across young, middle and older age groups were provided. All groups were normally distributed (W > 0.97, p > 0.05), and Mauchly’s test also indicated that the assumption of sphericity was not being violated (p > 0.05). For the four groups across two configurations, Table 1 lists the scores (Mean ± STD) of the four behavioral measures for the one-way ANOVA analysis. We did not find a significant (p < 0.05) difference between the groups (focal and non-focal) for the behaviors. These behavioral measures were included to evaluate if there are any association of focality with behavioral scores. A significant association could help the tDCS community to select subjects (based on that behavioral score) for whom the applied tDCS will optimally stimulate the target ROI. Our lack of significant differences in anxiety and depression scores for focal and non-focal groups could be due to the fact that our subjects were obtained from a database of healthy individuals. Future investigations could examine the relationship between such behavioral measures and focality in patient populations.

The focal ROIs were determined for all the regional volume parameters (CSF, GM and WM). A three-way ANOVA (performed on the groups obtained from montage F3-RSO at 1 mA) revealed significant differences (p < 0.05, Bonferroni corrected) in the regional CSF volume at the left caudate nucleus. There was a significant main effect of age [F (1, 2) = 85.84, p < 10⁻¹⁴] and an interaction effect of age and focality [F (1, 2) = 16.38, p < 0.0001] (Shown in red in Figure 1C). The post hoc comparison showed that the difference was significant (p < 0.0001) for the older group only, with subjects in the focal group showing higher volume (Figure 1C) than the non-focal group. The results remained similar for the tDCS configuration F3-RSO at 2 mA (not shown). The correlation between the current density at the left DLPFC (Target ROI) and the CSF volume at left caudate nucleus (focal ROI) was positive (∼ 0.16) and significant (p < 0.01) for both configurations (Figure 1D- 1 and 2 mA). The results of the White test (p = 0.12) confirmed that there is no heteroscedasticity. For both current doses (1 and 2 mA), the correlation of current density of left DLPFC with CSF volume of left caudate nucleus are calculated for the (i) Focal Group (r ∼ 0.28, p < 0.01) and (ii) non-focal (r ∼ 0.09, p > 0.05). A significant (p < 0.05) difference in the correlation coefficient between the two groups exists. The focal group with high current density at left DLPFC showed high CSF volume at left caudate nucleus.

Similar to previous studies (Gur et al., 1991; Lemaître et al., 2005; Mortamet et al., 2005; Lotze et al., 2019; Kijonka et al., 2020), significant effects of age and sex were observed in our analyses. Briefly, the significant (p < 0.05, Bonferroni corrected) main effect of age for the—(1) CSF volume was found in 66 ROIs (distributed throughout the brain), (2) WM was found in 9 ROIs (mainly in subcortical structures), and (3) GM was found in 11 ROIs (cortical and subcortical). The main effect of sex was found for the CSF volume only across the 4 ROIs (frontal and parietal areas). The other interaction effects across the combinations (age and sex, sex and focality, and age, sex and focality) were insignificant for the three regional volume parameters across the ROIs. Although significant effects of age and sex were observed in our analyses, we do not present the details of these effects here because age and sex had been investigated extensively in previous studies, and we do not wish to divert the attention from focality. However, as noted above, there was an age-focality interaction in the left caudate nucleus. We also performed a similar analysis after regressing out age and sex from the volumetric parameters. This did not alter the existing results, however, we found a significant [F(1, 2) = 72.34, p < 10⁻¹²) effect of age and an interaction effect of age and focality [F(1, 2) = 12.64, p < 0.001] in the regional CSF volume for the right caudate nucleus. The post hoc comparison also showed that the difference was only significant (p < 0.0001) for the older group. However, the correlation of the CSF volume at the right caudate nucleus with the current density at the target area (left DLPFC) was insignificant for both current doses (figure not shown).

The spread of the current density over an individual’s brain is shown in Figure 2A for the montage positioned at CP5-Cz with the target ROI at left IPL. Similar to the F3-RSO montage, the focal and non-focal group composition for the CP5-Cz configurations (1 and 2 mA) is in Table 2. The table shows the mean and standard deviation of DTDI for the two groups (focal and non-focal) across the two tDCS configurations (1 and 2 mA). The number of males and females in each group (focal/non-focal) is provided. The individuals in each group were distributed across young, middle, and older ages. The groups were normally distributed (W > 0.97, p > 0.05) and assumption of sphericity was not violated (p > 0.05). The scores for the four behaviors (MMSE, HADS_anxiety, HADS_depression, and PSQI) are also enlisted (Mean ± STD). The one-way ANOVA analysis of the scores of 4 behavioral measures for the two groups (focal and non-focal) did not find a significant difference between them.

Across the volume parameters, the two-way ANOVA analysis (performed on the groups obtained from 1 mA) revealed that the CSF volume at two focal ROIs, namely the left precentral [F(1, 1) = 10.15, p < 10⁻⁵] and postcentral gyrus [F(1, 1) = 18.76, p < 10⁻⁷] (shown with a black dotted line dividing the two red regions in Figure 2B) showed a significant main effect of focality (p < 0.05, Bonferroni corrected). The estimated marginal means found the non-focal group to have higher CSF volume compared to the focal group for both pre- and post-central gyrus (Figure 1C). The other volume parameters (WM or GM) were not significant. The result was also similar to the segregated groups for the current intensity of 2 mA (result not shown). A negative but significant (p < 0.01) correlation (r-value ranging from –0.15 to –0.18) was observed between the current density values at the target ROI (IPL) and the CSF volume of the focal ROIs (pre- and post-central gyrus) across the two current intensities (Figures 2D,E - 1 and 2 mA). The results of the White test (p-values ranging from 0.2306 to 0.2672) confirmed that there was no issue of heteroscedasticity in the correlation analysis across the two focal ROIs and two current doses. For both current doses (1 and 2 mA), the correlation of current density of IPL with CSF volume of pre- and post-central gyrus are calculated for the (i) Focal Group (r ∼ –0.33, p < 0.01) and (ii) non-focal (r ∼ –0.14, p < 0.01) groups. A significant (p < 0.05) difference in the correlation coefficient between the two groups exists. On a side note (as explained previously), the main effect of age and sex were similar for the two montages (CP5-Cz and F3-RSO). The interaction effects across the combinations (age and sex, age and focality, sex and focality, and age, sex and focality) were insignificant for all the volumetric parameters (regional CSF, WM, and GM).

Another interesting aspect that can be seen in both montages (F3-RSO and CP5-Cz, Figures 1, 2) across all the focal ROIs was that the distribution pattern of the current density does not appear to be the same for the two current doses (1 and 2 mA). This observation is in concordance with our previous findings, where we found a non-linear increase in current density with an increase in the current dose that is predominant in the older (> 60 years of age) population (Kashyap et al., 2021). This non-linearity has been reported in several other studies (Brunoni et al., 2012; Hsu et al., 2015; Habich et al., 2020) and might be the reason behind the observed differences in the pattern of distribution. This has been extensively discussed in our previous study (Kashyap et al., 2021); and we again report it here since the observation replicates.

For a tDCS configuration, we reduced the DTDI threshold in steps of 0.01. For the tDCS configuration F3-RSO at 1 mA, findings that were obtained with the predefined threshold of 0.75 were obtained until we have reduced the threshold to 0.70. For the tDCS configuration F3-RSO at 2 mA, the DTDI threshold could only be reduced to 0.73. Similarly, for the montage CP5-Cz at 1 and 2 mA, the DTDI can be reduced to 0.70 and 0.72, respectively. Altogether, the four tDCS configurations suggest that the threshold of 0.75 can be an acceptable milestone to gauge tDCS focality.