Stereo-EEG mapping of visual working memory with task-related high-gamma modulations

We describe a safe, informative, and easy-to-implement approach for presurgical mapping of visual working memory (VWM) with stereo-electroencephalography (SEEG).

Twenty-four patients with drug-resistant epilepsy, 11-23 years of age, performed a single-probe change detection VWM task, during SEEG monitoring. High-gamma modulations (HGMs) and connectivity modulations (weighted phase lag index) were computed for encoding, retention, recall, and response phases compared to a rest phase. Effects of lesioning sites with significant HGM on neuropsychological outcomes were analyzed.

Validity of our VWM task was evident by picture vocabulary scores covarying positively with increasing consistency and negatively with increasing maximum latency of the response times, respectively. VWM task elicited specific patterns of cortical HGMs during each phase: bilateral frontoparietal augmentation during encoding, replaced by widespread suppression during retention, followed by diffuse bilateral augmentation during the recall and response phases. Distinct neural circuits were also observed during each phase of the VWM task: left-lateralized phonological loop during encoding, short-range bilateral frontal connectivity during retention, and interhemispheric peri-Rolandic hubs during recall and response phases. Lesioning sites with significant high-gamma augmentation was associated with declines in domains directly relevant to VWM including working memory (slope -6.18, p < .001), visuospatial (-1.17, p = .031), picture vocabulary (-1.45, p < .001), and letter-word identification (-6.39, p = .004), whereas lesioning sites with significant high-gamma suppression, particularly in recall and response phases, affected domains sharing resources for neural computations with working memory, but not directly related to VWM including processing speed (-16.04, p = .049), calculation (-19.12, p = .020), and verbal comprehension (-9.09, p = .020).

Our observations extend theoretical models of VWM by providing phase-specific engagement of neural subsystems, particularly of phonological loop during encoding of even visually-cued stimuli and spatial rehearsal mechanisms during retention. Clinically, adverse neuropsychological outcomes from lesioning of VWM sites during epilepsy surgery support wider adoption of VWM mapping during presurgical evaluation of patients with drug-resistant epilepsy.