Multimodal imaging-based targeting approach for network-level brain stimulation

Neural network effects of transcranial direct current stimulation (tDCS) are poorly understood. Here, we introduce a prospective, empirically informed, multimodal functional magnetic resonance imaging (fMRI) framework for guiding target selection and hypothesis-based analysis in future focal tDCS-fMRI studies.

We illustrate our approach by using data of 37 healthy individuals (19 females; mean age ± SD = 25.8 ± 5.9) recruited from two tDCS-fMRI studies that were acquired at the same scanner and with placebo-tDCS. Participants completed two resting-state (RS) sessions and two task-fMRI sessions (object-location memory, OLM, or associative picture-pseudoword learning, APPL, experiments). Seed-based RS analysis identified functional networks originating from target regions for focal tDCS (right occipito-temporal cortex, rOTC; left ventral IFG, lvIFG) and established their test-retest reliability (TRR), using intraclass correlation coefficients (ICC). Dice coefficients quantified overlap between seeded RS networks and task-evoked activity to identify task-active regions potentially affected by downstream network effects from the target regions.

Seed-based analyses identified highly reliable ventral visual-limbic (rOTC) and language-related networks (lvIFG), with 72-77% of voxels showing good-to-excellent TRR (ICC ≥ 0.75). Only a subset of network voxels identified by the RS analyses overlapped with activity elicited by the experimental paradigms (ranging from 7.5-55%), with larger correspondence for the OLM (Dice: 0.249-0.349; APPL 0.065-0.106). Therefore, the degree of potential tDCS network effects varied substantially depending on the target region, the extent of its functional network and task-specific activity patterns. Degree of correspondence was further mediated by the selected contrasts-of-interest in the task-based analyses, with more conservative control conditions resulting in reduced overlap.

In sum, we established a principled multimodal fMRI framework bridging a critical gap in neuromodulation research. By integrating reliable intrinsic connectivity maps with task-evoked activity patterns, we provide a method to prospectively identify network-level targets for focal brain stimulation and generate hypotheses for tDCS-fMRI analyses. This approach shifts the rationale from stimulating isolated brain regions to strategically targeting key nodes within a predefined functional pathway.