Focality Oriented Selection of Current Dose for Transcranial Direct Current Stimulation

Rajan Kashyap, Sagarika Bhattacharjee, Ramaswamy Arumugam, Rose Dawn Bharath, Kaviraja Udupa, Kenichi Oishi, John E. Desmond, SH Annabel Chen, Cuntai Guan 1*# 1 School of Computer Science and Engineering, Nanyang Technological University, Singapore Psychology, School of Social Sciences (SSS), Nanyang Technological University, Singapore Centre for Research and Development in Learning (CRADLE), Nanyang Technological University, Singapore Department of Neuroimaging and Interventional Radiology, National Institute of Mental Health and Neurosciences, India Department of Neurophysiology, National Institute of Mental Health and Neurosciences, India The Johns Hopkins University School of Medicine, Baltimore, United States Lee Kong Chian School of Medicine (LKC Medicine), Nanyang Technological University, Singapore National Institute of Education, Nanyang Technological University, Singapore


Introduction
Transcranial Direct Current Stimulation (tDCS) is a noninvasive brain stimulation technique that could alleviate symptoms of several neurological and psychiatric brain disorders [1][2][3]. A conventional tDCS setup consists of anode and cathode placed over the scalp (referred as montage) with low intensity of current (~ 1-3mA) being injected to stimulate the target region of interest (ROI) [4,5]. However, the injected current gets diffused in the intermediary regions of the brain and might not essentially stimulate the target ROI with desired intensity [6,7]. Computational models that predict the pattern of current flow across the brain of an individual are used to optimize the tDCS stimulation parameters [8][9][10][11][12][13][14]. The amount of injected current (referred as 'current dose') plays an important role in the dispersal of stimulation intensity across the brain regions [15,16]. The distribution may vary from person to person and within a person based on the quantity of the dose [17][18][19].
Therefore, selection of optimal current dose for an individual's brain that could sufficiently stimulate the target ROI while minimizing the current in non-target ROIs is important [15,16].
In recent years there has been a growing interest towards individualization of current dose [15,16,20]. It has been reported that varying the current intensity at scalp for each individual can reduce the interindividual variability in the electric field intensity at the target ROI [20]. The current dose calculated through inverse modelling of tDCS induced electric field at the target ROI correlates with the motor thresholds generated by transcranial magnetic stimulation [15,16]. In a recent tDCS experiment using frontal montage and 2mA (fixed) current dose, individuals with high simulated current density at the target ROI (left dorsolateral prefrontal cortex) were found to have stronger improvements in working memory compared to those with low current density [21]. They also showed that individualizing the current dose by fixing the current density desired at the target region can maximize the benefits of tDCS [21]. Though the models are a step towards individualizing the current dose, they do not consider the spread of the field to intermediary (non-target) regions. The current flow in the intermediary regions have a vital role to play in determining the outcome of tDCS [6,12,[22][23][24][25]. It has been found that some brain regions may act as conduit clustering most of the current to a specific location that can deter the stimulation intensity expected at the target ROI [6,26]. Increasing the focality of stimulation in conventional tDCS setup has been an area of investigation [27][28][29][30]. Therefore, the approaches to individualize the current dose should consider the focality of stimulation in order to recommend the optimal intensity of input current.
In our previous work, we developed individual-Systematic-Approach-for-tDCS-Analysis (i-SATA) toolbox [31] that estimates the average current density received by target ROI and intermediary regions of an individual's brain after a montage has been simulated in Realistic-volumetric-Approach-based-Simulator-for-Transcranial-electric-stimulation (ROAST) toolbox [10]. We demonstrated the ease with which i-SATA toolbox can be applied on an individual brain to reverse calculate the current dose that can stimulate the target ROI with desired intensity [31]. This was done based on the assumption that electric field intensity at target ROI increases linearly with current dose by following the procedure laid down by Evans and colleagues [20] . Since we will be using it throughout the study, it will be helpful to familiarize our readers with an example. Suppose the calculated stimulation intensity at the target ROI is 0.25 mA/m 2 when 1 mA of current is applied on the scalp. To achieve an intensity of 0.5 mA/m 2  With i-SATA(MNI), we introduce the Dose-Target-Determination-Index (DTDI), a simple estimate that will quantify the focality of stimulation and facilitate the selection of optimal current dose required to stimulate the target ROI in an individual's brain. The index provides a comprehensive overview of the intensity of stimulation received by the target ROI and intermediary regions after a montage has been postprocessed in i-SATA(MNI). To explain DTDI, we will use the montage with anode positioned at F3 and cathode at right supra-orbital (RSO) (referred to as F3-RSO, Figure 1A). The montage has been shown to stimulate the left middle frontal gyrus [22,25] and is effective for depression [3,22,42] and working memory [43]. To make it easy for our readers to interpret how DTDI facilitates selection of the current dose, we will show the inter-individual as well as the intra-individual variation in the index by uniformly increasing the current dose. Finally, we will evaluate the variation in DTDI with age and sex of individuals. The purpose will be to explore if dose selection should be prioritised for any category (age and sex) of individuals.

Data
We obtained the T1-weighted (T1WI) magnetic resonance image (MRI) of the brain

Preprocessing with ROAST
We simulated the montage F3-RSO with the electrode size 5 × 5 cm 2 ( Figure 1A).  each location (voxel) is then used to obtain the average magnitude of current density received by each cortical region of the brain. This will provide an estimate of the current density induced in the target and intermediary region due to tDCS. As an example, we will postprocess the standard MNI 152 averaged head in i-SATA(MNI) for the three current doses (1mA, 2mA, and 3mA) using the F3-RSO montage to show the distribution of average current density across the cortical regions ( Figure 1B, C, D).

Dose Target Determination Index (DTDI)
The output of i-SATA (MNI) (i.e. the average current density in the target ROI and the non-target regions) is used to calculate the DTDI for a montage simulated at a current dose. For this, we will find the ROI that has the maximum value of average current density (peak region) amongst all the ROIs. DTDI is then calculated as = ℎ DTDI will lie in the range of 0 to 1. An ideal tDCS setup will expect the maximum intensity of stimulation (average current density) to be received at the target ROI, thereby generating a DTDI value equal to 1. However, the peak intensity may be received at nontargeted ROI. For an individual, the current dose for which DTDI is higher should be preferred over other doses. To make this clear, we will estimate the DTDI of three individuals across three current doses (figure 2). Hypothetically, the value of DTDI should remain constant across doses, since it is assumed that the current flow in the brain increases linearly with increase in current intensity [15,16,20,23,50].

Statistical Analysis of variation in DTDI
All individual MRIs were post processed in i-SATA(MNI) for the three current doses using the F3-RSO montage to estimate the DTDI's (Total = 90 MRI × 3 current doses = 270). We show the inter-and intra-individual variation in the DTDI for both sexes across the three age groups ( Figure 3). We performed three-way mixed ANOVA with age and sex as between subject and dose as within subject factor. Post-hoc analysis were performed to further characterize the nature of the main effects and interactions.
The package can be downloaded at (LINK_TO_BE_ADDED). A reference manual is also provided to aid users to run each step with ease.

Output of i-SATA(MNI) on the standard head model
The montage F3-RSO applied on the MNI 152 averaged head model simulated in ROAST is shown in figure 1A.  be seen that the current intensity at target-ROI is increasing but lesser number of regions are receiving current higher than the target ROI. As a result, the DTDI is increasing with increase in dose suggesting that higher current dose should be beneficial ( Figure 2B). Finally for the third individual, a decrease in DTDI is seen with increase in dose ( Figure 2C). The drop in DTDI from 1 mA to 2 and 3 mA seems to be due to increase in current in the right superior parietal lobule at 2 mA and 3 mA only. Although, the current intensity at target ROI is increasing with increase in dose but maximal amount of current is also getting dissipated to other brain regions. Thus, the conventional way of increasing the current dose to attain desired stimulation intensity at target ROI might result into stimulation of unwanted brain region (as seen for superior parietal lobule). For this individual, a lower dose showing higher DTDI can maximize the advantage of stimulation.

Statistical Analysis of variance in DTDI
Here we will highlight the change in DTDI with increase in dose for males and females across three age groups (Figure 3).  The inter-and intra-individual variation in DTDI clearly shows the current dose that could be appropriate for an individual to stimulate the target ROI after a montage has been fixed, and (B) showing the variation of DTDI for both sexes across the three age groups using three-way mixed ANOVA. The DTDI decreases with increase in age. In mid and older adults, females show higher focality compared to males for the three current doses (1mA, 2mA, and 3mA). In older adults only, the significant (p < 0.05, Bonferroni corrected) difference between DTDI at 1mA and 3 mA for both sexes conveys that higher current doses are required to appropriately stimulate the target ROI.

Discussion
In this paper, we extended the toolbox i-SATA [31] to the MNI reference space for users to obtain the average current density induced at each cortical region of an individual's brain due to tDCS. We then used these values to estimate DTDI as an objective measure to   Figure 1) and to an extent in young and middle aged individuals (Figure 2A and   3A). However, on the other hand, linearity appears to diminish with advancement in age ( Figure 2B, C and 3A) suggesting a potential non-linear relationship.
The different values of DTDI as a function of current dose across different subjects could be because the injected current might get clustered in brain areas (referred as hotspots), a phenomenon that has been widely reported in tDCS studies [6,25,26,[62][63][64]. Hotspots cause shunting of the current towards the surrounding brain tissue and a surge in the electric field strength at localised areas [65]. Areas that form hotspot can be away from the electrode site as well [65]. In the two cases presented in figure 2 (Subject 2 and 3), superior parietal lobule appears to be the hotspot. Here it is difficult to comprehend the neuroanatomical factors that attributes to the formation of such hotspots. It has been found that tissue heterogeneities and pathological alterations (like neurodegeneration and cerebral infarcts) are the primary contributors [26,65]. As we age, the atrophy in the neural configuration escalates the nonlinearities in the spatial distribution of induced electric field [66,67]. Care must be taken about possibilities of such hotspots for clinical application of tDCS [68]. DTDI that considers the current density in target and non-target areas inculcates the effect of hotspots to provide a rigorous estimate of optimal current dose. However, it is important to identify the factors that contribute to observed non-linearity and alterations in DTDI in future studies.
Since maximum stimulation might not be received at the target ROI and also not in a consistent location [6,17,18,69], the inter and intra-individual variation in DTDI can provide insights for appropriate determination of current dose based on age and sex of a healthy individual. In young adults, the focality of stimulation remains intact (approximately) across the doses ascertaining that there is flexibility in choosing (individualising) a dose depending on the extent of current density desired at the target ROI. However, the focality declines with advancing age (middle age onwards, see Figure 3). This decline is higher for males compared to females. Such sexual dimorphism in tDCS related effects have been reported in previous studies [70] and several factors related to cortical anatomy like volume [71], bone density [72], hormonal levels [72], and electrode location [73] have been postulated to account for it.
We also found that higher current dose can enhance the focality in older adults. This is in support of a recent study [74] that reported cerebral atrophy in older adults to cause the reduction in the amount of current reaching the target ROI. Altogether our findings suggest that determination of the current dose based on focality must be prioritised based on the age (> 40 years) and sex (especially males) of an individual.
We have shown the use of DTDI to titrate the current dose at the individual level. Finally, we would like to highlight that DTDI can be estimated from i-SATA as well.
However, simulation in i-SATA(MNI) is considerably faster than in i-SATA. This is because both i-SATA(MNI) and the integrated SPM anatomy toolbox for cortical labelling are MATLAB based and automated. This makes i-SATA(MNI) efficient to post-process large data sets, a trend that is emerging in neuroscientific research.

Conclusions
The study extends the i-SATA framework to the MNI atlas space. With i-SATA (MNI), it will be easier to calculate the individualized dose as suggested in previous studies [15,16,20,21]. Here we introduce the DTDI as measure to titrate the individualized current doses and select the optimum dose that has high focality and could appropriately stimulate the target ROI in an individual. Using a montage that has been found to be optimal for DLPFC stimulation, DTDI analysis across a broad spectrum of men and women of different age groups revealed that focality decreases with advancing age, especially in males with more than 40 years of age. Finally, the study reveals that selection of current dose that increases the focality is strictly necessary for older (> 60 years) individuals irrespective of sex.