Evaluating personalized rTMS impact on network functional connectivity via precision methods

65 

Abstract Submission 

Anne Billot¹, Randy Buckner¹, Stephanie McMains², Mark Eldaief¹

¹Harvard University, Cambridge, MA, ²Boston University, Boston, MA

Anne Billot  
Harvard University
Cambridge, MA

Randy Buckner  
Harvard University
Cambridge, MA

Stephanie McMains, PhD  
Boston University
Boston, MA

Mark Eldaief, MD  
Harvard University
Cambridge, MA

Repetitive transcranial magnetic stimulation (rTMS) is used as a treatment for neuropsychiatric disorders, such as depression (O'Reardon et al., 2007). Clinically, rTMS targets are typically identified using anatomical landmarks (Cash et al., 2020). However, multiple studies have shown that the efficacy of rTMS treatments depends on its ability to modulate specific functional networks (Liston et al., 2014). Limited evidence exists regarding the impacts that targeting specific networks has on the functional connectivity of different circuits. Moreover, evidence has focused on network estimates that are based on group data. Recent work in our lab and others (Braga & Buckner 2017) have used precision functional connectivity MRI (fcMRI) estimates to delineate functional network topography in the individual. This study evaluates the specificity of personalized rTMS upon network-level connectivity, defined with precision fcMRI methods, at the individual level.

Two healthy adults (S1 and S2) underwent a baseline scan (3T MRI) to collect individualized functional data in order to perform fcMRI analyses to delineate the location of their salience (SAL) network. After the baseline session, each individual completed 30 rTMS sessions, at least 4 days apart from each other. Ten sessions targeted a representation of the SAL network in the left dorsolateral prefrontal cortex (LDLPFC), ten sessions targeted the SAL in the right DLPFC (RDLPFC), and 10 sessions targeted the same LDLPFC site but with sham rTMS, administered in a counterbalanced order. RTMS sessions were directly followed by a resting-state MRI scan that included at least two BOLD runs of fixation. After quality control, 59 and 53 runs were used for each subject, respectively. BOLD fMRI data were acquired using a multiband gradient-echo echo-planar pulse sequence. A T1 scan was acquired using an MPRAGE sequence (see acquisition parameters in Braga et al., 2019). MRI data were analyzed using FSL, Freesurfer, SPM, and custom, in-house software. The SAL target was derived using a seed map encompassing the entire SAL network, defined by Yeo et al., 2011, but excluding a DLPFC mask in each hemisphere. The specific target was chosen as the region exhibiting the maximal functional connectivity with the SAL seed map. RTMS was administered with a Magventure Cool B65 A/P liquid-cooled coil, capable of active and sham stimulation, with the following parameters: 20Hz stimulation, at 110% of the subject's resting motor threshold over 45 trains (2s and 40 pulses per train), with an intertrain interval of 28s for a total of 1800 pulses (22.5 min) (Eldaief et al., 2023). During stimulation, a neuronavigation system was used to stimulate the predefined targets precisely and reproducibly across sessions by loading the subject's fcMRI data, overlaid on the subject's native-space structural MRI.

To measure differences in post-TMS correlation strengths across conditions, we first used all 31 fMRI runs to estimate 15 functional networks using precision MRI estimates through a Multi-Session Hierarchical Bayesian Model (MS-HBM) (Braga and Buckner 2017), and to determine the network identity of the stimulated targets in each subject. Then, for each rTMS condition, we used the first runs of the ten MRI sessions to compute correlation strength between the target ROI and each functional network using the mean Fisher-transformed Pearson's correlation coefficient between the time course of all vertices within each target ROI (left and right) and the vertices within each predefined network (excluding the target region).

Precision mapping of fcMRI showed that the main network targeted was CG-OP on both sides in S1, and SAL/PMN on the left and DN-A on the right in S2 (Fig1). Connectivity results showed opposite rTMS effects within the targeted left CG-OP in S1 and right DN-A in S2 (Fig2).

Future research will prospectively use precision fcMRI mapping to investigate individual rTMS effects on distinct circuits.

Non-invasive Magnetic/TMS ¹

Connectivity (eg. functional, effective, structural) ²

Task-Independent and Resting-State Analysis

Anatomy and Functional Systems

Cortical Anatomy and Brain Mapping

ADULTS

Cognition

Cortex

FUNCTIONAL MRI

MRI

Psychiatric Disorders

Transcranial Magnetic Stimulation (TMS)

Other - precision mapping