A comprehensive approach for TMS targeting identifies better targets than typical approaches.
Poster No:
117
Submission Type:
Abstract Submission
Authors:
Dylan Nielson1, Marie Zelenina2, Safa Rahman2, Andre Zugman3, Daniel Pine4, Francisco Pereira2
Institutions:
1National Institute of Mental Health, Washington, DC, 2National Institute of Mental Health, Bethesda, MD, 3NIMH, Bethesda, MD, 4National Institute of Mental Health (NIMH), National Institutes of Health (NIH), Bethesda, MD
First Author:
Co-Author(s):
Daniel Pine
National Institute of Mental Health (NIMH), National Institutes of Health (NIH)
Bethesda, MD
National Institute of Mental Health (NIMH), National Institutes of Health (NIH)
Bethesda, MD
Introduction:
Functional connectivity is being used clinically to target TMS treatment for depression (1,2). Most targeting approaches restrict their search space by first identifying a cortical target based on functional connectivity, then optimize the stimulator position to reach that target without consideration of any additional cortical regions that might be stimulated (2,3). In fact, stimulator positions found this way sometimes deliver stimulation to a broad swath of the cortex that is at least as strong as that delivered to the cortical target (Fig. 1A). The dispersed patterns of stimulation delivered by TMS are a challenge for delivering stimulation to specific, compact cortical targets.
Here we propose a comprehensive approach for TMS targeting of depression in which the stimulated cortical area of all reasonable stimulator orientations are simulated. We then select the stimulator position with the strongest mean anticorrelation with the subgenual cortex (SGC), as opposed to the stimulator position that does the best job at stimulating the single most anticorrelated cluster.
Here we propose a comprehensive approach for TMS targeting of depression in which the stimulated cortical area of all reasonable stimulator orientations are simulated. We then select the stimulator position with the strongest mean anticorrelation with the subgenual cortex (SGC), as opposed to the stimulator position that does the best job at stimulating the single most anticorrelated cluster.
Methods:
METER Sample
The NIMH Multi-Echo TEst-Retest sample consists of multi-session data from 7 healthy adults collected on a 3-T GE MR-750 (4) (Fig. 1B). We collected 0.8 mm isotropic T1- and T2-weighted images and echo planar resting state sequences (2.5mm isotropic, TR=2.5 s; TEs=[12.9 ms, 32.2 ms, 51.6 ms, 70.9 ms]). This data was collected under protocol 01-M-0192 approved by the NIH IRB.
Targeting approach
Our targeting approach was inspired by Lynch et al.'s (5) approach for stimulating functional networks and makes use of their preprocessing pipeline (6) (detailed in Fig. 1C), Amongst other changes, we have refined the search of stimulator positions to focus on delivery of stimulation to the gyral lip, since modeling indicates neurons here have the lowest activation threshold (7).
Target Quality Evaluation
We evaluate target quality based on uncertainty weighted activation probability maps. We expect that stimulation with a higher weighted mean SGC correlation will be more likely to have a clinical effect. It is also possible that the sign of the stimulated cortex is relevant to clinical effects, so we evaluated the weighted proportion of stimulated cortex anticorrelated with the SGC.
Target Reliability Evaluation
Many studies of targeting reliability have reported intersession distance between cortical targets, but this is a poor metric if there are multiple nearly equivalent stimulation sites. To account for this, we evaluate the quality of the targets from each session with the connectivity data from the other. This cross-session analysis tells us if a target remains a good quality target across sessions, even if it is not necessarily the single best target in all sessions.
Code is available at https://github.com/nih-fmrif/contarg.
The NIMH Multi-Echo TEst-Retest sample consists of multi-session data from 7 healthy adults collected on a 3-T GE MR-750 (4) (Fig. 1B). We collected 0.8 mm isotropic T1- and T2-weighted images and echo planar resting state sequences (2.5mm isotropic, TR=2.5 s; TEs=[12.9 ms, 32.2 ms, 51.6 ms, 70.9 ms]). This data was collected under protocol 01-M-0192 approved by the NIH IRB.
Targeting approach
Our targeting approach was inspired by Lynch et al.'s (5) approach for stimulating functional networks and makes use of their preprocessing pipeline (6) (detailed in Fig. 1C), Amongst other changes, we have refined the search of stimulator positions to focus on delivery of stimulation to the gyral lip, since modeling indicates neurons here have the lowest activation threshold (7).
Target Quality Evaluation
We evaluate target quality based on uncertainty weighted activation probability maps. We expect that stimulation with a higher weighted mean SGC correlation will be more likely to have a clinical effect. It is also possible that the sign of the stimulated cortex is relevant to clinical effects, so we evaluated the weighted proportion of stimulated cortex anticorrelated with the SGC.
Target Reliability Evaluation
Many studies of targeting reliability have reported intersession distance between cortical targets, but this is a poor metric if there are multiple nearly equivalent stimulation sites. To account for this, we evaluate the quality of the targets from each session with the connectivity data from the other. This cross-session analysis tells us if a target remains a good quality target across sessions, even if it is not necessarily the single best target in all sessions.
Code is available at https://github.com/nih-fmrif/contarg.
Results:
The Comprehensive approach stimulates cortical areas with a stronger mean anticorrelation with the SGC (p = 7.25x 10-7) and a greater proportion of anticorrelated cortical area (p = 3.39 x 10-7) than the Restricted method (Fig 2A). It was also more reliable that the Restricted approach with stronger mean anticorrelation with the SGC (p = 7.30x 10-5) and a greater proportion of anticorrelated cortical area (p = 3.99 x 10-5) in the crossed-session evaluation (Fig 2B). 4 of 7 participants had the same target in both sessions with the Comprehensive method, but none did with the Restricted method (Fig 2C).
Conclusions:
In this pilot, we showed that taking a comprehensive approach to TMS targeting allows simulation to be delivered to an area of the cortex that is more anticorrelated with the SGC than a restricted targeting approach. The length of resting state sequences we collected allows us to evaluate the quality of the targeting but it may exaggerate measures of target reliability compared to shorter sequences. However, this work does demonstrate the feasibility and promise of a comprehensive TMS targeting approach.
Brain Stimulation:
Non-invasive Magnetic/TMS 2
TMS 1
Modeling and Analysis Methods:
Task-Independent and Resting-State Analysis
Keywords:
Open Data
Open-Source Code
Transcranial Magnetic Stimulation (TMS)
Workflows
Other - Functional Connectivity
1|2Indicates the priority used for review
Provide references using author date format
2. Cole EJ et al. (2022): Stanford Neuromodulation Therapy (SNT): A Double-Blind Randomized Controlled Trial. American Journal of Psychiatry 179: 132–141.
3. Cash RFH et al. (2021): Personalized connectivity-guided DLPFC-TMS for depression: Advancing computational feasibility, precision and reproducibility. Human Brain Mapping 42: 4155–4172.
4. Nielson DM et al. (2023): NIMH METeR (Multi-Echo Test-Retest). Openneuro. https://doi.org/10.18112/OPENNEURO.DS004787.V1.1.0
5. Lynch CJ et al. (2022): Automated optimization of TMS coil placement for personalized functional network engagement. Neuron 110: 3263-3277.e4.
6. Lynch CJ et al. (2020): Rapid Precision Functional Mapping of Individuals Using Multi-Echo fMRI. Cell Reports 33: 108540.
7. Aberra AS et al. (2020): Simulation of transcranial magnetic stimulation in head model with morphologically-realistic cortical neurons. Brain Stimulation 13: 175–189.