Spatially constrained ICA from a language task predicts Sensorimotor Network in patients with tumors

Background Although resting-state fMRI is a promising alternative to task-fMRI in presurgical mapping, protocols to localize simultaneously and accurately both language and motor areas are still missing. Purpose To propose a methodology based on the connectivity analysis to identify motor and language areas using a single 6-minute fMRI language task in presurgical patients with space-occupying lesions. Methods In a retrospective study and using a language fMRI task (verb generation), we established limits of the motor cortex in 40 presurgical patients. Single-subject spatially-constrained ICA was performed on verb generation scans to extract somatomotor network. Sensitivity, and specificity between predicted SMN and hand motor task were calculated. Variables effect was analyzed through ANOVA. Results Forty patients (mean age, 40.50 ± 13.99 [standard deviation]; 21 men) diagnosed with a space-occupying lesion were included. Somatomotor network extracted from spatially-constrained ICA and language lateralization from task-fMRI were identified in 40/40 (100%) patients. Using the motor task as reference standard, ipsilesional voxel-to-voxel mean sensitivity, and specificity for spatially-constrainted ICA at voxel level was 78%, and 75%, respectively. No significant differences were found between ipsilesional and contralesional hemispheres in sensitivity or specificity. Right contralesional hemisphere, higher scanner field strength, and typical language lateralization, showed higher sensitivity values. Conclusions Our study demonstrates the possibility of identifying somatomotor network connectivity from a language task while determining language lateralization in patients with space-occupying lesions. Our technique is a promising alternative to reduce scanning time and cost as only one fMRI sequence is needed to determine the two main eloquent functions.