Surface imaging for real-time patient respiratory function assessment in intensive care.

Abstract

Monitoring of physiological parameters is a major concern in Intensive Care Units (ICU) given their role in the assessment of vital organ function. Within this context, one issue is the lack of efficient noncontact techniques for respiratory monitoring. In this paper, we present a novel noncontact solution for real-time respiratory monitoring and function assessment of ICU patients. The proposed system uses a Time-of-Flight depth sensor to analyze the patient's chest wall morphological changes in order to estimate multiple respiratory function parameters. The automatic detection of the patient's torso is also proposed using a deep neural network model trained on the COCO dataset. The evaluation of the proposed system was performed on a mannequin and on 16 mechanically ventilated patients (a total of 216 recordings) admitted in the ICU of the Brest University Hospital. The estimation of respiratory parameters (respiratory rate and tidal volume) showed high correlation with the reference method (r=0.99; P<0.001 and r=0.99; P<0.001) in the mannequin recordings and (r=0.95, P<0.001 and r=0.90, P<0.001) for patients. This study describes and evaluates a novel noncontact monitoring system suitable for continuous monitoring of key respiratory parameters for disease assessment of critically ill patients.