Classification of Level of Consciousness in a Neurological ICU Using Physiological Data.

Area of Science:

• Critical care medicine
• Neuroscience
• Biomedical engineering

Background:

• Impaired consciousness is common in intensive care unit (ICU) patients, impacting care and prognosis.
• Current methods lack continuous monitoring of consciousness levels and alerts for changes.
• Investigating physiological signals for real-time consciousness classification in critically ill patients.

Purpose of the Study:

• To explore the utility of physiological signals for classifying consciousness levels in ICU patients.
• To develop a method for continuous monitoring and alerting of consciousness changes.
• To improve patient care and outcomes through objective consciousness assessment.

Main Methods:

• Retrospective observational study of prospectively collected data from 61 subarachnoid hemorrhage (SAH) and 178 intracerebral hemorrhage (ICH) patients.
• Classification of consciousness levels (comatose, VS/UWS, MCS-, command following) based on neurological examination.
• Utilized time-series features from physiological signals and extreme gradient boosting for classification across different variable sets (minimal, common ICU, extended neurological ICU).

Main Results:

• Achieved an AUROC of 0.72 for VS/UWS vs. MCS- classification in SAH patients using the extended set.
• AUROC for VS/UWS vs. MCS- in ICH patients reached 0.64 using the minimal set.
• Demonstrated the feasibility of classifying consciousness states using physiological signals with varying performance based on data subsets.

Conclusions:

• Physiological signals show potential for classifying consciousness states in ICU patients.
• Further development could enable alarm systems for timely detection of consciousness changes.
• Enhanced monitoring may significantly improve patient management and clinical outcomes in critical care settings.