Real-Time Mapping of Natural Speech in Children with Drug-Resistant Epilepsy

Abstract

Epilepsy surgery is a recognized treatment option in patients with drug-resistant epilepsy (DRE). The pre-surgical evaluation in DRE involves determination of anatomical and functional relationships between the seizure-onset zone and cortical areas sub-serving essential sensorimotor and language functions. Language mapping in clinical practice at present, involves electrical stimulation of pairs of implanted subdural electrodes using increasing current strengths, as the patient repetitively performs a task related to language production and/or reception. However, this procedure of electrical cortical stimulation (ECS) is fraught with multiple safety issues and concerns about ecological validity of the chosen tasks. In view of the challenges associated with ECS, alternative methods of language mapping have been explored including measurement of stimulus-locked decrease in power in α band of EEG (event-related desynchronization) and more importantly, event-related synchronization in the high-γ (≥70 Hz) frequency band. Early studies using stimulus-locked high-gamma synchronization have used conventional language tasks like picture naming. Subsequent studies have described the topography of power changes related to natural conversation between patient and family/staff. These studies documented the topography of such power changes to correspond with activation observed with formal language tasks and also with conventional anatomy of language network. However, these studies used offline analysis of recorded signals after the data collection. In this manuscript, we describe real-time visualization of activation associated with natural speech. We expect this technology to allow real-time dynamic mapping of human language networks, which can potentially improve the safety and access of epilepsy surgery, as well as enhance the knowledge about developmental organization of language networks in human brain.