Resting State fMRI for Motor Cortex Mapping in Children with Epilepsy

Task-based fMRI is commonly used to localize motor functions prior to epilepsy surgery. Not all children are able to follow task instructions in the scanner, limiting fMRI motor mapping. We present a data-driven method to automatically delineate the motor cortex using task-free, resting state fMRI (rsfMRI) data of children with epilepsy and controls. We preprocessed anatomical and rsfMRI images (64 patients, 12 controls) using fMRIPrep. We used FSL MELODIC tool for whole brain ICA estimation in multiple orders, to generate several sets of independent components. We defined a motor component selection process based on Discriminability Index-based Component Identification (DICI) score, resulting from comparing binarized component to the Human Motor Area Template (HMAT) in subject's native brain space. Components with the largest DICI score and their peak in the template were combined to form the whole-brain ICA Motor Map (wIMM). We validated wIMM by comparing individual results with finger tapping motor task activation, and evaluated its reproducibility in controls with two runs of rsfMRI. The hit rate between wIMM and motor task activation ranged 60 to 77% across all participants. Sensitivity of wIMM for capturing the task activation peak was 87.5% among 32 patients and 100% in 11 controls with available task results. The mean Dice similarity in results from repeated runs was 0.70 in controls. Our results shows the sensitivity and reproducibility of an automated motor mapping based on ICA analysis of rsfMRI in children with epilepsy. The ICA maps may provide different, but useful, information than task fMRI. Future studies will expand our method to mapping other brain functions, and may lead to a surgical planning tool for patients who cannot perform task fMRI and help predict their postsurgical function.