EEG synchronization measures are early outcome predictors in comatose patients after cardiac arrest

Abstract

Quantitative methods might increase the prognostic yield of currently used multi-modal approaches.

Auditory discrimination improvement predicts awakening of postanoxic comatose patients treated with targeted temperature management at 36 °C

“…These markers are robustly related to poor neurological recovery (see [5] for an overview). By contrast, their presence does not predict whether the patient will recover from coma -with the exception of EEG background reactivity [6,7], which however showed imperfect inter-rater agreement despite attempts for standardized interpretations (e.g. [8,9]).…”

Electroencephalography-based power spectra allow coma outcome prediction within 24 h of cardiac arrest

“…by quantifying the evoked response from auditory stimuli 8,9 or by characterising features of the neural activity at rest. 10 In this new study, we capitalize on evidence that disorders of consciousness are associated to functional impairments of cortico-thalamic connections and associated with specific EEG power spectra features. 11 To the best of our knowledge, no studies provided quantitative estimations of EEG power spectra predictive performance for long-term outcome during early stages of coma.…”

EEG‐based outcome prediction after cardiac arrest with convolutional neural networks: Performance and visualization of discriminative features

“…The performance of t‐VGG was comparable or exceeded that of other quantitative or clinical approaches previously applied to (smaller, but more homogeneous) parts of the same cohort. For instance, a Bayesian classifier based on eight predetermined quantitative EEG features (various bivariate synchronization measures) reached a AUC of 0.81 (Zubler et al, ). A clinical multimodal prognostication approach combining visual EEG features, somatosensory evoked potentials, brainstem reflexes and neuron specific enolase also reached an AUC of 0.81 for predicting unfavorable outcome (Tsetsou, Novy, Pfeiffer, Oddo, & Rossetti, ).…”

“…The quality of each EEG, especially the presence of electromyographic artifacts, was assessed post hoc by a certified electroencephalographer (F.Z.) as described in Zubler et al ().…”

“…Several EEG patterns have been shown to be associated with either unfavorable outcome (e.g., suppressed background, burst‐suppression ‐ especially if showing identical bursts, epileptiform activity) or favorable outcome (continuous background, background reactivity; Hofmeijer et al, ; Westhall et al, ). Quantitative (computer‐based) methods have been proposed, either to perform as surrogate electroencephalographers (Tjepkema‐Cloostermans et al, ), to increase precision and speed of interpretation (Ruijter, Hofmeijer, Tjepkema‐Cloostermans, & van Putten, ; Ruijter, van Putten, & Hofmeijer, ; Rundgren, Westhall, Cronberg, Rosén, & Friberg, ), or to detect EEG features not easily recognizable by human eye (Beudel, Tjepkema‐Cloostermans, Boersma, & van Putten, ; Pfeiffer et al, ; Zubler et al, ). Most quantitative approaches are “feature engineered,” meaning that an algorithm was explicitly designed to detect or quantify predefined features of the EEG signal such as amplitude, frequency spectrum, presence of spiky elements, and linear or nonlinear interactions between the channels (Zubler, Bandarabadi, Kurmann, Gast, & Schindler, ).…”