Integration of video-based AVPU consciousness detection in emergency and hospital care: Consc.ia: Product development of a video-based AVPU assessment platform for emergency and hospital care

Abstract

Accurate and timely assessment of patient consciousness is essential in emergency medicine, guiding triage, care escalation, and patient safety. However, documentation is often inconsistent across pre-hospital and hospital settings due to operational pressure, differing workflows, and the subjective nature of observational scoring. The AVPU scale (Alert, Verbal, Pain, Unresponsive) is widely used for rapid assessment, yet it is often under-recorded or documented retrospectively, which reduces its value for decision-making and continuity of care.This thesis presents Consc.ia, a product-oriented decision support solution designed to assist healthcare professionals by automating AVPU inference from video and enabling interoperable documentation across emergency and hospital workflows. The proposed approach combines a video-based inference pipeline with a semi-automated validation layer, allowing AI-generated AVPU predictions to be reviewed and confirmed by clinicians before storage and reporting. To support feasibility and future deployment, the system is designed around key real-world requirements, including real-time processing, privacy-by-design, workflow compatibility, and structured integration through HL7 FHIR resources for downstream EHR use.A dedicated dataset was developed for proof-of-concept evaluation, comprising 136 recordings from 58 participants representing simulated AVPU assessment procedures under controlled conditions. The system architecture and deployment scenarios are discussed for Emergency Medical Services, emergency departments (EDs), and in-hospital monitoring, with attention to interoperability, governance, and operational constraints. An adoption-oriented analysis is also introduced to support the transition from prototype to product.Overall, this work outlines a practical pathway for developing video-based consciousness assessment solutions that improve documentation quality, reduce variability, and support digital transformation in emergency care.