Improving presurgical language mapping by a method for optimally sorting independent components of resting-state fMRI

Pre-surgical planning often involves task-based functional magnetic resonance imaging (fMRI) in the context of intractable epilepsy or brain tumors. Resting-state fMRI can be used for the same goal, with the advantage of being a simpler technique that does not require the patient to cooperate in complex cognitive tasks. However, the methods for resting-state fMRI analysis are not yet robust or of practical usage. This work proposes an algorithm for sorting components resulting from independent component analysis (ICA) that emphasizes the language resting-state network. We recruited 20 healthy volunteers and acquired resting-state and task-based fMRI using three linguistic tasks. Task data was processed using general linear model analysis, while resting-state networks were extracted using ICA. An automated IC sorting procedure was developed based on three characteristics: spatial similarity with a probability map, low/high frequency ratio, and IC reliability over several bootstrapping folds. Task-related activation consistent with the language network was identified at the subject-specific level. The algorithm is shown to sort ICs with the resting-state language maps appearing among the first three with an accuracy of 74%. Overall, the Dice coefficient showed a good overlap between the sorted ICs of relevance and the task language maps. Results showed that resting-state networks were more specific and less sensitive than task-based maps. We expect that the proposed algorithm for optimal sorting will contribute towards making ICA usage viable in the clinical context and become a reliable alternative method for pre-surgical planning.