Abstracts

Rationale: Task-free, resting state fMRI (rsfMRI) can identify functional brain networks and may lead to alternative or confirmatory pre-surgical brain mapping tools. An independent component analysis (ICA)-based mapping method has shown promise in motor mapping for epilepsy patients [1]. Here we used data from the Epilepsy Connectome Project (ECP) to investigate the intra-subject reproducibility of ICA motor mapping. We hypothesized that the method is equally reproducible in patients and controls.


Methods: From ECP, T1-weighted and rsfMRI data from 54 patients with temporal lobe epilepsy and 41 controls were pre-processed using fMRIPrep, where each subject had 2-6 rsfMRI runs. We used an ICA mapping method similar to our previous work [1]. Briefly, rsfMRI images were denoised and decomposed into spatially independent components utilizing FSL MELODIC. Using an iterative template-matching process based on Discriminability Index-based Component Identification (DICI) compared to a Human Motor Area Template (HMAT [2]), a final ICA z-map was generated and thresholded at 2 for each rsfMRI run. We calculated the average Dice Similarity Coefficient across multiple runs from the same subject and in comparison to HMAT.


Results: There were no significant differences between the two groups in their mean within-subject Dice values (Controls: 0.30 (SD=0.10) vs. Patients: 0.26(0.08), T-test p >0.05; Cohen’s D=0.35), and mean subject similarity to HMAT (Controls: 0.28 (0.05) vs. Patients: 0.27 (0.06, p >0.05; Cohen’s D=0.203).


Conclusions: Our ICA motor mapping method showed reproducibility at an individual level for patients and controls alike despite potential functional differences. While within-subject similarity (or map overlap) varied at the tested threshold, comparable subject-to-template similarity ranges in both groups support the robustness of our motor mapping technique and its potential in its applicability in clinical and research settings.


Funding: SS and AM are undergraduate students at UCSD. We thank Dr. Jeffrey Binder and the Epilepsy Connectome project for making their dataset available. Authors report no funding or conflict of interest related to this study.



References
[1] Krishnamurthy, M., et al. J Neuroimaging 2022, 32:1201-10.

[2] Mayka MA, et al. Neuroimage 2006; 31:1453-74.



Figure 1: Violin plots of control and patient group Dice Similarity Scores. Left panel shows the average similarity between multiple runs of individual subjects, not significant (NS) between the two groups. Right panel shows the distribution of similarities between motor maps and HMAT, the template used to select the ICA components.