Methodology for tDCS integration with fMRI

Abstract

Understanding and reducing variability of response to transcranial direct current stimulation (tDCS) requires measuring what factors predetermine sensitivity to tDCS and tracking individual response to tDCS. Human trials, animal models, and computational models suggest structural traits and functional states of neural systems are the major sources of this variance. There are 118 published tDCS studies (up to October 1, 2018) that used fMRI as a proxy measure of neural activation to answer mechanistic, predictive, and localization questions about how brain activity is modulated by tDCS. FMRI can potentially contribute as: a measure of cognitive state-level variance in baseline brain activation before tDCS; inform the design of stimulation montages that aim to target functional networks during specific tasks; and act as an outcome measure of functional response to tDCS. In this systematic review, we explore methodological parameter space of tDCS integration with fMRI spanning: (a) fMRI timing relative to tDCS (pre, post, concurrent); (b) study design (parallel, crossover); (c) control condition (sham, active control); (d) number of tDCS sessions; (e) number of follow up scans; (f) stimulation dose and combination with task; (g) functional imaging sequence (BOLD, ASL, resting); and (h) additional behavioral (cognitive, clinical) or quantitative (neurophysiological, biomarker) measurements. Existing tDCS-fMRI literature shows little replication across these permutations; few studies used comparable study designs. Here, we use a representative sample study with both task and resting state fMRI before and after tDCS in a crossover design to discuss methodological confounds. We further outline how computational models of current flow should be combined with imaging data to understand sources of variability. Through the representative sample study, we demonstrate how modeling and imaging methodology can be integrated for individualized analysis. Finally, we discuss the importance of conducting tDCS-fMRI with stimulation equipment certified as safe to use inside the MR scanner, and of correcting for image artifacts caused by tDCS. tDCS-fMRI can address important questions on the functional mechanisms of tDCS action (e.g., target engagement) and has the potential to support enhancement of behavioral interventions, provided studies are designed rationally.

Keywords: brain stimulation; methodology; neuroimaging; tDCS-fMRI integration.

Conflict of interest statement

CUNY has patents on brain stimulation with M. Bikson as inventor. M. Bikson has equity in Soterix Medical Inc. and serves on the scientific advisory boards of Boston Scientific and GlaxoSmithKline. The other authors have potential conflict of interest to be disclosed.

© 2019 The Authors. Human Brain Mapping published by Wiley Periodicals, Inc.

Figures

Figure 1

Search strategy for inclusion, exclusion criteria. To identify relevant studies, we searched PubMed using logical combination of keywords “tDCS” OR “transcranial direct current stimulation”) AND (“functional magnetic resonance imaging” OR “fMRI” OR “functional MRI” OR “fcMRI” OR “functional connectivity MRI” OR “rsfMRI” OR “resting‐state MRI”

Figure 2

Permutations in trial design for integration of MRI‐based neuroimaging methods with tDCS. (a) fMRI timing can occur pre‐ and/or post‐stimulation outside scanner (sequential‐outside scanner approach), pre‐ and/or post‐stimulation inside scanner (sequential‐inside scanner approach) or concurrently during stimulation to evaluate tDCS effect on ongoing neural activities. (b) Study designs include single arm open label studies with no control condition, parallel approach where subjects are randomly assigned to either active or sham group and crossover approach, where subjects participate in both active and sham sessions with a random order. (c) Control conditions consist of sham, active control or both. (d) tDCS may be applied over multiple sessions with an expectation of cumulative (time) effect. (e) Imaging may include multiple follow up sessions to evaluate after effects longitudinally. (f) Stimulation dose (i.e., electrode montage, stimulation duration, and intensity) and combination with task will determine outcomes. (g) Essentially any functional imaging sequence can be utilized to evaluate stimulation effects depending on the study questions/hypothesis. (h) tDCS‐imaging study designs can combine a multitude of other objective or subjective assessments concurrent or combined with imaging and/or tDCS

Figure 3

Functional imaging sequences and analysis methods used in the literature of tDCS‐fMRI studies

Figure 4

An experiment design and self‐report (subjective) results in a double blind, crossover, and sham controlled study. (a) Each subject underwent a counterbalanced study design with two stimulation types (real and sham) on two separate days with a one‐week wash out period. Functional MRI (fMRI) data were acquired before and after each stimulation session. Subjects rated their immediate methamphetamine craving before and after both stimulation and imaging sessions on a visual analogue scale (VAS), with a score range from 0 to 100, where 0 and 100 indicated “no craving” and “extreme craving,” respectively. (b) Self‐report results for four different timepoints within the experiment. Subjects rated their craving level before and after each functional MR scan and stimulation session (i.e., sham, active). (b.3) tDCS significantly reduced craving compared to sham (p < .05). (b.4). Study includes 15 subjects (colored circles represent each subject)

Figure 5

Three major roles functional MRI can play in integration with tDCS. (1) Montage selection studies: functional imaging could contribute to montage selection to localize stimulation to target functionally activated brain areas. (2) Prediction studies: baseline fMRI could provide baseline measures of neural network activation and inform the interpretation of later behavioral/neural responses to tDCS (i.e., “Baseline fMRI as Predictor”). (3) Mechanistic studies: utilizing fMRI to investigate brain changes underlying tDCS effects and ultimately determine where, when and how stimulation affects brain function and associated behavior

Figure 6

Different MR‐compatible stimulator setups inside the scanner. (a) Setup comprises of two filter boxes (inside scanner room (MR‐compatible) and outside scanner room (not MR‐compatible)) and cables come from the control room through RF waveguide tube. (b) Setup includes an RF filter adapter that connects to the RF penetration panel (the device will have the same ground as the RF shield of MRI room)

Figure 7

Induced cortical electric field magnitude for the montages considered. From left to right: 3D surfaces of skin and brain, cross‐sectional coronal and sagittal images. OPEN: Enough gel for satisfactory impedance in the interface between electrode and skin but gel is contained in HD cups. SHORT: Excessive gel smeared between adjacent electrodes reflecting excessive gel application and/or headgear motion. This results in a highly conductive path for current to flow above the surface of the skin (“short circuit”) that reduces current penetration to the brain

Figure 8

Effect of current‐induced stray magnet field on Bz in watermelon. Electrodes (white arrows) are positioned on left and right side along the equator (orientation visible in proton density MRI in [b]). MGE data from +/−2 mA current injection was used to calculate effective Bz (+4 mA) for cables oriented along z (c) and ~25° from z (d). Cables immersed in agar (red arrows) provide contrast in the T1 underlays of the Bz images (c), (d). After experiments, the watermelon was sliced and photographed (a) at the level of (b), showing the structural sources of heterogeneity in (b)

Figure 9

Concurrent stimulation & imaging artifact. fMRI activation during hand motor task (a) (finger tapping) in a healthy subject with tDCS sponge electrodes placed over hand motor region. Clearly, the rectangular “deactivation” seen as a green vertical line on the scalp (b), and that overlaps exactly with the position of the cathode, is an artifact

Figure 10

Workflow of computational modeling integration with functional MRI. Given the two separate maps of standard space used in fMRI and head models used in modeling, in the first step, high resolution EF map in fine meshed geometry should be interpolated to voxel‐wise standard space. Then, both head models and functional images should be registered to standard space to enable comparison

Figure 11

Correlation of electric field with neural activation change (post–pre) in the study introduced in Figure 4. (a) Atlas based segmentation. Regions include: frontal pole (FP), superior frontal gyrus (SFG), medial frontal gyrus (MFG), frontal orbital cortex (FOC), and frontal medial cortex (FMC). (b) Current flow‐based segmentation. Areas are selected based on current flow pattern generated based on bilateral DLPFC montage, 5 × 7 electrodes. Regions of interest include: superior frontal gyrus (SFG), ventrolateral prefrontal cortex (vlPFC), medial orbitofrontal cortex (mOFC), and frontal pole (FP). Blue circles: sham stimulation, red circles: active stimulation