IntelliLung Takes the Next Step Towards AI-Supported Mechanical Ventilation
Mechanical ventilation is a life-saving treatment for critically ill patients, but finding the right ventilator settings remains a complex challenge. Inappropriate settings can cause additional lung damage, while patients’ needs may change rapidly during treatment. The EU-funded IntelliLung project aims to address this challenge by developing an AI-based decision support system that provides personalized recommendations for mechanical ventilation. A new publication in BMJ Open, led by researchers at TUD Dresden University of Technology, presents the study protocol for an international clinical validation involving 530 patients across four intensive care units in Germany, Spain and Poland. The study marks an important step towards evaluating the technology under real clinical conditions and lays the groundwork for future research into its potential benefits for patients.
“We want to help clinicians tailor mechanical ventilation to each patient’s needs and make treatment safer for critically ill patients. With this study, we are taking an important step towards that goal by evaluating how AI-generated recommendations compare with clinical decisions in real intensive care settings. The findings will help us understand the potential of AI to support clinicians in making complex treatment decisions,” said Dr. Raphael Theilen, first author of the publication and trial coordinator at TUD Dresden and Dresden University Hospital.
From AI Algorithms to Real-World Clinical Validation
The IntelliLung system uses reinforcement learning, a machine learning approach, to analyze patient-specific information and recommend key ventilator settings, including oxygen concentration, breathing rate, airway pressure and tidal volume. The algorithm was developed using historical clinical data from more than 12,500 intensive care stays. In the ongoing prospective observational study, the system processes real-time data from hospital information systems and medical devices through a dedicated, secure IT infrastructure. The IntelliLung Team will compare its recommendations with the settings chosen by the ICU clinicians to assess their agreement and identify potential differences in clinical decision-making. Importantly, the AI recommendations remain hidden from the treating teams and do not influence patient care at this stage. The findings will provide a foundation for future interventional trials investigating whether AI-supported ventilation can improve patient safety and clinical outcomes. The study is therefore the next important step towards bringing the IntelliLung system to the patient bedside.
Combining Clinical Expertise and Technological Innovation
The study brings together clinical partners from four European intensive care units: University Hospital Carl Gustav Carus Dresden in Germany, Parc Taulí University Hospital in Sabadell and Hospital Universitario de La Princesa in Madrid, Spain, as well as the National Medical Institute in Warsaw, Poland. Led by the Principal Coordinating Investigator Jakob Wittenstein and the team at TU Dresden, the study builds on close collaboration between intensive care specialists, AI researchers, software developers and medical technology experts. While clinicians contribute their expertise in mechanical ventilation and patient care, technical partners develop the AI algorithms, secure data infrastructure and integration of medical devices and hospital systems. This interdisciplinary approach is essential to ensure that IntelliLung addresses the practical needs and challenges of everyday clinical practice and is technically robust.