Clinical evaluation of an AI-based prototype for contactless respiratory monitoring in children

Background: The gold standard for respiratory assessment, PSG (polysomnography), requires a multitude of sensors to be attached directly to the patient. Especially in children, this procedure regularly causes discomfort, impairs sleep quality and sometimes leads to rejection of sensors.

Objective: The aim of our study was the clinical evaluation of a digital camera prototype designed for contactless detection and AI-based assessment of breathing movements in children.

Methods: The design, development and preclinical validation of the infrared camera system prototype was conducted in a previous study. The core principle of the system is the acquisition of a four-dimensional spatial image over several hours and a subsequent AI-based analysis of the region-of-interest, which allows the analysis of movements in the millimeter range. Using this approach, we recorded and analyzed the respiratory movements of 27 children (aged 6 months to 12 years) during sleep. At the same time, a standard PSG was recorded.

Results: In summary, we recorded and analyzed approximately 300 hours of nocturnal breathing in different sleeping positions and settings. For the present study, breathing frequency was chosen as the primary outcome parameter. To ensure comparability for the analysis, data was taken 5 minutes after entry into the first deep sleep phase (N3). Regarding the determination of respiratory rate, we found an almost perfect correlation between PSG and the contactless prototype camera system (r2 = .95).

Conclusion: The present study demonstrates the feasibility of non-contact respiratory detection using a newly developed 4D camera prototype in children.