Targeting the pathological network: Feasibility of network-based optimization of transcranial magnetic stimulation coil placement for treatment of psychiatric disorders

Abstract

It has been recognized that the efficacy of TMS-based modulation may depend on the network profile of the stimulated regions throughout the brain. However, what profile of this stimulation network optimally benefits treatment outcomes is yet to be addressed. The answer to the question is crucial for informing network-based optimization of stimulation parameters, such as coil placement, in TMS treatments. In this study, we aimed to investigate the feasibility of taking a disease-specific network as the target of stimulation network for guiding individualized coil placement in TMS treatments. We present here a novel network-based model for TMS targeting of the pathological network. First, combining E-field modeling and resting-state functional connectivity, stimulation networks were modeled from locations and orientations of the TMS coil. Second, the spatial anti-correlation between the stimulation network and the pathological network of a given disease was hypothesized to predict the treatment outcome. The proposed model was validated to predict treatment efficacy from the position and orientation of TMS coils in two depression cohorts and one schizophrenia cohort with auditory verbal hallucinations. We further demonstrate the utility of the proposed model in guiding individualized TMS treatment for psychiatric disorders. In this proof-of-concept study, we demonstrated the feasibility of the novel network-based targeting strategy that uses the whole-brain, system-level abnormity of a specific psychiatric disease as a target. Results based on empirical data suggest that the strategy may potentially be utilized to identify individualized coil parameters for maximal therapeutic effects.

Keywords: brain network; electric field calculation; individualized treatment; psychiatric disorder; transcranial magnetic stimulation.

FIGURE 1

Schematic illustration of network targeting model. (A) Stimulation network. For transcranial magnetic stimulation (TMS) administrated with a given combination of parameters (i), the generated E-filed (ii) defines direct TMS effects on the local cortical region. Group-level rsFC (iii) provides a visualization of the functional network affected via the stimulated cortical region, i.e., the stimulation network (iv). (B) Stimulation networks vary among individuals due to both the coil setting and geometry and productivity of individuals’ intra-cranial tissues. (C) Comparing to the pathological network of a given disease, (D) stimulation networks showing spatial anti-correlation are hypothesized to be associated with better clinical improvement induced by TMS (Fox et al., 2014).

FIGURE 2

Network targeting model predicts the equation-based transcranial magnetic stimulation (TMS) treatment efficacy at empirical dorsolateral prefrontal cortex (DLPFC) sites in a large depression cohort. (A) Empirical target sites of major depressive disorder (MDD) are shown in MNI-152 (Fonov et al., 2011). (B) Restoration of TMS parameters from targeted cortical sites. (C) Network targeting accuracy (NTA) of empirical sites across different individuals, each represented with a colored dot (N = 68). (D) Correlation between the average NTA and the equation-based HDRS total improvement (p = 9.32 × 10^–6, one-tailed).

FIGURE 3

Network targeting accuracy predicts treatment efficacy in the clinical major depressive disorder (MDD) cohort. (A) Coil placement of left prefrontal cortex (PFC) patients shown on individual head models. (B) Correlation between network targeting accuracy (NTA) and Montgomery–Asberg depression rating scale (MADRS) total improvement (N = 27, p = 0.043, one-tailed).

FIGURE 4

Network targeting accuracy predicts treatment efficacy in the clinical cohort of schizophrenia with auditory verbal hallucinations (AVH). (A) Coil placement of active group patients shown on individual head models. (B) Correlation between network targeting accuracy (NTA) and auditory hallucination rating scale (AHRS) total improvement (N = 15, p = 0.016, one-tailed).

FIGURE 5

Major depressive disorder (MDD) simulation experiment. (A) Illustration of positions and orientations of a representative individual. Large black dots represent 125 positions in the search space. For each position, 12 coil orientations, in the normal plane at the position (0°∼−165°, 15-degree intervals), were tested. Network targeting accuracy (NTA) was calculated for each pair of position and orientation. (B) NTA value distribution in the search grid. Each position in the 2-D grid represents a combination of position and orientation. (C) Maximum NTA was found in all patients (yellow border). Search space was interpolated from 125 × 12 to 27,977 × 12 for visualization purposes. (D) The optimal transcranial magnetic stimulation (TMS) coil placements are shown in individual scalp spaces. The cyan arrow represents 0° at each position.

FIGURE 6

Schizophrenia with auditory verbal hallucinations (AVH) simulation experiment. (A) Illustration of positions and orientations of a representative individual. Large black dots represent the 122 positions in the search space. For each position, 12 coil orientations (0°∼−165°, 15-degree intervals) were tested. Network targeting accuracy (NTA) was calculated for each pair of position and orientation. (B) NTA value distribution in the search grid. Each position in the 2-D grid represents a combination of position and orientation. (C) Maximum NTA found in all patients (yellow border). Search space was interpolated from 122 × 12 to 58,470 × 12 for visualization purposes. (D) The optimal transcranial magnetic stimulation (TMS) coil placements are shown in individual scalp spaces. The cyan arrow represents 0° at each position.