A Touch-Free Guardian Angel in the Incubator: Hungarian Breakthrough Transforms the Monitoring of Preterm Infants

Monitoring the healthy long-term development of preterm infants is an exceptionally complex task in neonatal intensive care units. In addition to continuously tracking vital signs such as breathing, heart rate and body temperature, it is equally important to monitor infants’ spontaneous movements, posture and behavioural patterns in a consistent and objective manner.

Spontaneous motor activity is a direct indicator of neurological maturity and overall health in preterm infants. Movement patterns can help clinicians identify possible neurological abnormalities. However, manual visual observation is time-consuming and subjective, and cannot realistically be carried out around the clock.

A team from the Computational Optical Sensing and Processing Laboratory at the HUN-REN Institute for Computer Science and Control (HUN-REN SZTAKI) has developed an innovative solution to this problem. Led by Research Fellow Péter Földesy, the engineers worked with clinical experts from the Department of Neonatology at Semmelweis University, including Professor Miklós Szabó, to create a camera-based monitoring system powered by artificial intelligence.

For the study, the researchers monitored 88 infants in incubators and analysed 8,400 hours of video footage. They used the resulting data to develop a new model that could eventually be applied in the care of preterm infants.

The system described in the recently published study analyses images from cameras positioned above the incubators in real time. Deep-learning algorithms trained by researchers at HUN-REN SZTAKI can identify an infant’s posture with centimetre-level accuracy — determining, for example, whether the baby is lying in a supine, prone or lateral position (on their back, stomach or side). The system can also distinguish between small, coordinated movements and sudden reactions that may indicate stress.

The Hungarian-developed AI system continuously records and organises the data, providing nurses and doctors with immediate and reliable feedback.

The information collected by the system can also provide a more accurate picture of each infant’s individual sleep–wake cycle. Aligning treatment and nursing procedures with babies’ natural biological rhythms has been shown to support their development, reduce stress and potentially shorten the length of their hospital stay.