VUNO (338220.KQ) announced Wednesday that it has published research results on an artificial intelligence-based early warning system that predicts the need for mechanical ventilator intubation in pediatric critical care patients in the international journal Heart & Lung.
The study was conducted as a joint research project between VUNO and research teams from the Department of Pediatrics and Department of Pulmonology at Pusan National University Yangsan Hospital. The research team analyzed data from pediatric intensive care unit patients to develop a deep learning model that predicts the need for invasive mechanical ventilation intubation and compared it with existing models.
Acute respiratory failure is one of the major factors that can lead to pediatric ICU admission and cardiac arrest. As delays in assessing patient conditions increase risk levels, timely invasive mechanical ventilation intubation is critical. However, pediatric patients span a wide age range with diverse underlying conditions, making it difficult to detect deterioration early. In particular, most existing prediction models have been developed primarily for adults, limiting their applicability to pediatric patients.
To address this, the research team analyzed 1,318 electronic medical records of patients under 18 years old admitted to the pediatric ICU at Pusan National University Yangsan Hospital from 2012 to 2022, developing and validating an AI-based early warning system called DeePedIMV that predicts the need for invasive mechanical ventilation up to 8 hours in advance. Unlike existing models limited to specific disease groups, DeePedIMV reflects acute deterioration from various causes, enabling universal prediction.
The study results showed that DeePedIMV recorded prediction accuracy (AUROC) of approximately 0.88, demonstrating excellent predictive performance. This performance was consistent across all groups regardless of age and disease type, with the highest accuracy recorded in patients aged 1 year or younger. The system also recorded approximately 0.47 on AUPRC, an indicator that precisely identifies actual risk situations, proving performance more than three times higher than existing models. Additionally, by reducing alarm frequency to less than half at the same sensitivity level, the system can improve workflow efficiency for medical staff.
"This research has demonstrated that the deep learning algorithm we developed can reduce unnecessary alarms while proactively predicting high-risk pediatric critical care patients who require invasive mechanical ventilation intubation," said Joo Sung-hoon, Chief Technology Officer of VUNO. "We will continue our efforts to ensure VUNO's technology contributes to patient safety not only for adults but also for pediatric patients."