Neural Response to Familiar Names Predicts Outcome of Comatose ICU Patients: A Prospective Observational Cohort Study

2026-06-04 · Source: Machine learning : nature.com subject feeds · Field: Health & Wellbeing — Clinical Care & Medical Practice, Medical Devices & Health Technology, Artificial Intelligence & Machine Learning · Depth: Expert, short

Summary

A prospective observational cohort study involving 89 comatose patients from five intensive care units demonstrated that neural responses to familiar names can predict patient outcomes. Researchers utilized a state-of-the-art EEG frequency tagging approach, presenting patients with rapid streams of familiar names and unintelligible control sounds. EEG responses tracking familiar names were extracted in the frequency domain and positively correlated with Glasgow Outcome Scale-Extended (GOSE) scores, assessed at 1, 3, and 6 months post-injury. A machine learning model, integrating these EEG responses with clinical characteristics, achieved strong predictive performance. The model yielded AUCs of 0.86, 0.88, and 0.86 in the test set, and 0.91, 0.90, and 0.85 in an external validation set, for predicting outcomes at 1, 3, and 6 months, respectively. These findings indicate that residual processing of familiar names, detectable via EEG, holds potential for prognostic assessment in comatose ICU patients.

Key takeaway

For neurologists and intensivists assessing comatose patients, integrating EEG-based familiar name processing into your prognostic toolkit could significantly enhance outcome prediction. This method offers objective insights into residual cognitive function, complementing traditional clinical assessments. You should consider implementing frequency tagging EEG to detect covert cognition, potentially informing critical care decisions and family counseling regarding long-term recovery prospects.

Key insights

EEG responses to familiar names predict outcomes in comatose ICU patients.

Principles

• Covert cognition detection aids prognosis.
• EEG frequency tagging reveals neural processing.
• Combining EEG with clinical data improves prediction.

Method

Patients were presented with rapid streams of familiar names and control sounds. EEG responses were extracted in the frequency domain using frequency tagging, then integrated with clinical data into a machine learning model for outcome prediction.

In practice

• Use EEG frequency tagging for covert cognition.
• Integrate neural markers with clinical data.
• Assess patient outcomes at 1, 3, and 6 months.

Topics

• Comatose Patient Prognosis
• Electroencephalography
• Frequency Tagging
• Covert Cognition
• Machine Learning Models
• Glasgow Outcome Scale-Extended

Best for: AI Scientist, Research Scientist, Domain Expert

Related on AIssential

• DeepTokenEEG Enhancing Mild Cognitive Impairment and Alzheimers Classification via Tokenized EEG Features — DeepTokenEEG offers a lightweight, high-accuracy EEG-based method for Alzheimer's disease detection using spatial and temporal tokenization.
• RealICU: Do LLM Agents Understand Long-Context ICU Data? A Benchmark Beyond Behavior Imitation — RealICU is a new benchmark for LLMs in ICU settings, using hindsight-annotated data to overcome limitations of historical clinician actions.
• Stringology-Based Motif Discovery from EEG Signals: an ADHD Case Study — Stringology-based motif discovery offers a novel, complementary approach to analyze EEG dynamics in neurodevelopmental disorders like ADHD.
• Deep Sleep Classification via EEG Signal Criticality: A Passive BCI Approach for Sleep-Improvement Neurofeedback — High-accuracy passive BCI sensing is achievable via probabilistic decoding of EEG criticality.
• PhysioMotion Artifact: A task-driven EEG dataset with point-wise motion artifact annotations — PhysioMotion Artifact is a new EEG dataset with detailed point-wise annotations for 14 types of physiological motion artifacts.
• End-to-End Machine Learning for Depressive State Classification via EEG and fNIRS — Biological signal analysis via EEG and fNIRS offers an objective framework for depressive state classification.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.