Pervasive Vision-Based System for Simultaneous Physiological Parameter Tracking in Intensive Care Units: A Dual-Site Clinical Evaluation

Critique Open Research & Review

Abstract

Continuous monitoring of physiological parameters in intensive care units (ICUs) is essential for early detection of clinical deterioration, prevention of adverse events, and optimization of patient management. Conventional monitoring systems rely primarily on contact-based sensors such as electrocardiography leads, pulse oximeters, and respiratory belts, which may cause discomfort, restrict patient mobility, and increase the risk of infection or skin injury during prolonged use. Recent advances in computer vision, biomedical signal processing, and artificial intelligence have enabled the development of camera-based contactless monitoring systems capable of measuring multiple vital signs simultaneously. However, the clinical reliability, scalability, and real-world applicability of such systems in critical care environments remain insufficiently validated. This study presents a pervasive vision-based monitoring platform designed for simultaneous extraction of multiple physiological parameters, including heart rate, respiratory rate, oxygen saturation, and motion-related indicators, using non-contact imaging sensors in intensive care settings. The proposed system integrates multi-wavelength imaging, remote photoplethysmography, motion analysis, and machine learning–based signal reconstruction to achieve continuous monitoring without physical attachment to the patient. A dual-site clinical evaluation was conducted across two independent hospital ICUs to assess the robustness, accuracy, and clinical usability of the system under real-world conditions. The methodological framework combines advanced video-based physiological measurement techniques with adaptive filtering, signal quality assessment, and intelligent feature extraction to ensure reliable operation in complex ICU environments characterized by varying lighting conditions, patient movement, and clinical interventions. Performance was evaluated against standard bedside monitoring equipment using statistical agreement analysis, error metrics, and event detection capability. Results demonstrate that the proposed system achieves clinically acceptable accuracy for multiple vital parameters while significantly improving patient comfort and reducing sensor-related complications. The dual-center validation confirms the reproducibility of performance across different clinical infrastructures and patient populations. The findings support the feasibility of deploying pervasive vision-based monitoring as a complementary or alternative solution to conventional contact-based systems in critical care. This research contributes to the advancement of non-contact medical monitoring technologies and provides evidence for their integration into next-generation intelligent ICU environments.

Authors 3

1. Lukas Schneider

   Technical University of Munich

   Affiliation as printed

   Department of Computer Science and Biomedical Engineering , Technical University of Munich , Munich , Germany

2. Hanna Voge

   Heidelberg University

   Affiliation as printed

   Institute of Medical Informatics , Heidelberg University , Heidelberg , Germany

3. Markus Reinhardt Aachen Department of Electrical Engineering and Information Technology

   RWTH Aachen University

   Affiliation as printed

   Department of Electrical Engineering and Information Technology , RWTH Aachen University , Aachen , Germany