Peter W. Kaplan

Purpose: Recent research on quantitative EEG in coma has proposed several metrics correlating with consciousness level. However, the heterogeneous nature of coma can challenge the generalizability of these measures. This study investigates alpha-coma, an electroclinical pattern characterized by a widespread, nonreactive alpha rhythm often linked to poor outcomes. The aim was to quantify the electrophysiological features of alpha-coma and compare them to the alpha rhythm in awake controls, seeking clearer insights into quantitative EEG analysis in comatose states. Methods: Fourteen alpha-coma patients were retrospectively selected from University Hospitals of Geneva and age-matched with 14 healthy control subjects from an open-source dataset. EEG data were preprocessed and analyzed to extract power spectra, spectral decay (aperiodic activity), sample entropy, and functional connectivity. Results: Alpha-coma patients did not differ in alpha power but exhibited significantly higher levels of spectral decay ( p < 0.001), suggesting a convergence toward an inhibitory state. Sample entropy was significantly higher in alpha-coma patients ( p = 0.01), indicating an increase in the cortical complexity in alpha-coma compared with healthy subjects. Conclusions: Alpha-coma shows increased aperiodic activity and EEG complexity, despite similar alpha power and clustering coefficient. The increased aperiodic activity aligns with findings in other comatose patients, including those sedated or with subcortical dysfunction. However, the increased entropy contradicts existing literature, suggesting that alpha-coma may represent a state of widespread cortical dysfunction likely resulting from nonhierarchical, turbulent brain activity. This indicates that the loss of consciousness does not guarantee consistent cortical measures across the whole spectrum of EEG patterns.

In intensive care units (ICUs), critically ill patients are monitored with electroencephalography (EEG) to prevent serious brain injury. EEG monitoring is constrained by clinician availability, and EEG interpretation can be subjective and prone to interobserver variability. Automated deep-learning systems for EEG could reduce human bias and accelerate the diagnostic process. However, existing uninterpretable (black-box) deep-learning models are untrustworthy, difficult to troubleshoot, and lack accountability in real-world applications, leading to a lack of both trust and adoption by clinicians. We developed an interpretable deep-learning system that accurately classifies six patterns of potentially harmful EEG activity - seizure, lateralized periodic discharges (LPDs), generalized periodic discharges (GPDs), lateralized rhythmic delta activity (LRDA), generalized rhythmic delta activity (GRDA), and other patterns - while providing faithful case-based explanations of its predictions. The model was trained on 50,697 total 50-second continuous EEG samples collected from 2711 patients in the ICU between July 2006 and March 2020 at Massachusetts General Hospital. EEG samples were labeled as one of the six EEG patterns by 124 domain experts and trained annotators. To evaluate the model, we asked eight medical professionals with relevant backgrounds to classify 100 EEG samples into the six pattern categories - once with and once without artificial intelligence (AI) assistance - and we assessed the assistive power of this interpretable system by comparing the diagnostic accuracy of the two methods. The model's discriminatory performance was evaluated with area under the receiver-operating characteristic curve (AUROC) and area under the precision-recall curve. The model's interpretability was measured with task-specific neighborhood agreement statistics that interrogated the similarities of samples and features. In a separate analysis, the latent space of the neural network was visualized by using dimension reduction techniques to examine whether the ictal-interictal injury continuum hypothesis, which asserts that seizures and seizure-like patterns of brain activity lie along a spectrum, is supported by data. The performance of all users significantly improved when provided with AI assistance. Mean user diagnostic accuracy improved from 47 to 71% (P<0.04). The model achieved AUROCs of 0.87, 0.93, 0.96, 0.92, 0.93, and 0.80 for the classes seizure, LPD, GPD, LRDA, GRDA, and other patterns, respectively. This performance was significantly higher than that of a corresponding uninterpretable black-box model (with P<0.0001). Videos traversing the ictal-interictal injury manifold from dimension reduction (a two-dimensional representation of the original high-dimensional feature space) give insight into the layout of EEG patterns within the network's latent space and illuminate relationships between EEG patterns that were previously hypothesized but had not yet been shown explicitly. These results indicate that the ictal-interictal injury continuum hypothesis is supported by data. Users showed significant pattern classification accuracy improvement with the assistance of this interpretable deep-learning model. The interpretable design facilitates effective human-AI collaboration; this system may improve diagnosis and patient care in clinical settings. The model may also provide a better understanding of how EEG patterns relate to each other along the ictal-interictal injury continuum. (Funded by the National Science Foundation, National Institutes of Health, and others.).

In the case of suspicion of nonconvulsive status epilepticus (NCSE), reactivity on electroencephalograms (EEGs) can provide valuable diagnostic information. Reactivity refers to responses to auditory or somatosensory stimulation, with changes in amplitude and frequency of background activity. Because of self‐perpetuating processes and the failure of self‐terminating mechanisms, status epilepticus is unlikely to cease when patients spontaneously move, and it cannot typically be stopped by external stimulation (i.e., auditory and tactile stimuli). The defining EEG characteristic of absence status epilepticus is the presence of bilateral, synchronous, symmetric, rhythmic paroxysmal activity that shows little or no reactivity to sensory stimulation. On the other hand, in metabolic/toxic or multifactorial encephalopathies, triphasic waves (TWs) are influenced by the level of vigilance. TWs may be transiently abolished when patients increase their level of alertness from a drowsy/lethargic state to a state of wakefulness. This reactivity is only observed when patients can be aroused by a somatosensory or auditory stimulus. This reactivity tends to disappear with increasing severity of the disease and in comatose patients. In patients without preexisting developmental and epileptic encephalopathy, this pattern of stimulus‐induced wakefulness with transient improvement of the EEG is a major criterion in determining that the EEG patterns are not ictal. This criterion of reactivity on EEGs, beyond the classical clinical/EEG criteria of NCSE (Salzburg criteria), should now be systematically added.