When the target is a moving target: Practical issues in using task fMRI for rTMS targeting

912 

Abstract Submission 

Lysianne Beynel¹, Bruce Luber¹, Hannah Gura², Zeynab Rezaee¹, Ekaete Ekpo¹, Zhi Deng¹, Janet Joseph³, Paul Taylor¹, Sarah Lisanby¹

¹National Institute of Health, Bethesda, MD, ²University of Pennsylvania, Philadelphia, PA, ³Brown University, Providence, RI

Lysianne Beynel  
National Institute of Health
Bethesda, MD

Bruce Luber  
National Institute of Health
Bethesda, MD

Hannah Gura  
University of Pennsylvania
Philadelphia, PA

Zeynab Rezaee  
National Institute of Health
Bethesda, MD

Ekaete Ekpo  
National Institute of Health
Bethesda, MD

Zhi-De Deng, PhD  
National Institute of Health
Bethesda, MD

Janet Joseph  
Brown University
Providence, RI

Paul Taylor  
National Institute of Health
Bethesda, MD

Sarah Lisanby  
National Institute of Health
Bethesda, MD

Individualized task-based fMRI is often used to define rTMS targets. However, determining the optimal number of trials to attain sufficient statistical power while balancing the effects of practice, learning, and fatigue remains a challenge. This choice can substantially impact target selection and thus, rTMS effects. In this study, participants performed a numerical Stroop task (NST). In each trial, a pair of numbers, one larger in font size than the other, was displayed and participants were asked to select the number with the higher numerical value. Trials were randomly and evenly distributed across two conditions: Congruent (the visually larger number was numerically larger), and Incongruent (the visually larger number was numerically smaller). rTMS was applied to the spot within the right intraparietal sulcus (IPS) showing the strongest activation in the Incongruent > Congruent contrast across all blocks. No significant rTMS effect was observed on task performance. To evaluate whether the lack of effect could be due to a change in activation across blocks (resulting from practice, learning, and/or fatigue), we investigated how activations differed between the first and last pairs of blocks.

Nineteen healthy adults (14 female, 5 male; mean age 39 ± 14 years old) were enrolled, and then practiced a block of 48 trials of the NST. They then performed four more blocks of 48 trials each while in the scanner. fMRI data were analyzed with AFNI (3). First, @SSwarper was used (4) for skull stripping and alignment of the anatomical image from native to MNI space. Then afni.proc.py was used to setup a full pipeline for the fMRI analysis. The hemodynamic response was modeled with a gamma function using the onset of the Congruent and Incongruent trials. In this first follow-up analysis, the Incongruent > Congruent contrast for all trials was computed. In a second analysis, activations in the Incongruent > Congruent contrast were compared between the first and second halves of the acquisition.

The analysis of all trials revealed moderate activation in the left IPS (t > 2.10, 23 voxels), along with some larger clusters in subcortical structures, including the bilateral corpus callosum, and the left cingulate cortex, known to be involved in conflict monitoring (1). Noticeable changes were found in activation pattern when comparing the first and last pairs of NST blocks. The comparison between the Incongruent > Congruent contrast obtained for the last blocks vs. the first ones revealed a strong negative cluster (t > 2.16, p < 0.05, 1827 voxels, Figure 1a) largely spanning the bilateral IPS. Interestingly, we also observed a small positive cluster in the left dorsolateral prefrontal cortex. IPS activations observed in the first blocks (Figure 1b) totally disappeared in the last blocks (Figure 1c). At the behavioral level, accuracy remained quite high and constant across all blocks while reaction time continued to improve (Figure 2), suggesting that fatigue did not play a role and that in fact task efficiency continued to become optimized.

While more trials should provide more power (2) for rTMS target selection, this analysis indicated that the observed measures in the last blocks differed from those earlier and highly impacted the overall activation estimations. Since behavioral performance kept improving, a likely interpretation is of a learning effect during which one (or both) of two changes occurred: either the IPS was highly involved in the earlier learning stages of the task and became less necessary with practice, with possible shifts to other regions; or the number of IPS neurons necessary for task performance decreased with the optimization caused by practice, with subsequent diminishment of signal. This suggests that defining a rTMS target requires a better understanding of the local and network changes associated with task practice and learning effects to estimate the best parameters.

Transcranial Magnetic Stimulation (TMS)

Cognition

FUNCTIONAL MRI

Learning

Executive Function, Cognitive Control and Decision Making ¹

Activation (eg. BOLD task-fMRI) ²