Artificial intelligence brings personalized ventilation closer to reality in ARDS

Acute respiratory distress syndrome (ARDS) remains a major challenge in critical care, with mortality rates often exceeding 35–40%. Mechanical ventilation is essential for survival but can itself cause further lung injury if not optimally adjusted. Current ventilation protocols are based on population averages and often fail to reflect the dynamic physiology of individual patients.

In a new perspective article, a team of researchers from Hospital General Universitario Gregorio Marañón and Universidad Rey Juan Carlos (Madrid, Spain) has reviewed recent developments in the use of artificial intelligence (AI) to regulate mechanical ventilation for patients with ARDS. Their work describes how AI can bridge this gap between patients’ needs and ventilation protocols by continuously analyzing multiple data sources—ventilator waveforms, blood gases, vital signs, and imaging—to guide real-time, patient-specific ventilator adjustments. This work was made available online on December 2, 2025, in the Journal of Intensive Medicine.

“Our goal is to move from one-size-fits-all ventilation to precision support guided by AI,” explains Dr. Javier Muñoz, the study’s lead author and Head of the ICU at Hospital General Universitario Gregorio Marañón. “AI systems can process hundreds of physiological variables per second, anticipate complications such as asynchrony or oxygenation failure, and support clinical decisions that keep the patient’s lungs within a safe mechanical range.”

The paper highlights several technological breakthroughs:

• Machine learning algorithms (including convolutional and recurrent neural networks) trained on high-resolution ventilator signals now outperform conventional scoring systems in predicting readiness for weaning.
• Reinforcement learning models can simulate optimal ventilator settings to maximize ventilator-free days while reducing oxygen toxicity.
• Explainable AI (XAI) and hybrid physiological models help ensure that algorithmic recommendations remain interpretable, safe, and clinically traceable.
• Electrical impedance tomography (EIT) integrated with AI can generate bedside maps of lung aeration to guide positive end-expiratory pressure (PEEP) titration.

The authors emphasize that ARDS is the ideal testbed for AI-driven ventilation because of its complexity, heterogeneity, and high mortality. Integrating biological and physiological “subphenotypes” into AI systems may allow clinicians to tailor ventilation strategies to each patient’s underlying response type—for example, distinguishing hyperinflammatory from hypoinflammatory profiles. (See Figure)

However, the study also underscores the barriers to implementation: the need for multicenter datasets, regulatory oversight, and robust validation across different ventilator brands and patient populations. Ethical safeguards and “fallback” mechanisms are equally important to maintain clinician control.

“AI should not replace the clinician,” Dr. Muñoz notes. “It should act as an intelligent co-pilot—supporting decisions, ensuring safety, and learning from every breath.”

According to the authors, the next five years will likely see AI tools embedded directly into ICU ventilators, capable of recommending safe settings, predicting complications hours in advance, and improving synchrony between patient and machine.

The study concludes that combining AI with physiological principles could open a new era of truly personalized mechanical ventilation, potentially reducing lung injury, shortening ventilation duration, and improving outcomes in ARDS worldwide.

***

Reference
DOI: 10.1016/j.jointm.2025.10.003

About Dr. Javier Muñoz
Dr. Javier Muñoz is Head of the Intensive Care Unit at Hospital General Universitario Gregorio Marañón and a researcher at the Instituto de Investigación Sanitaria Gregorio Marañón in Madrid, Spain. He is an intensive care specialist with more than three decades of clinical and research experience in critical care, including work on extracorporeal membrane oxygenation (ECMO), acute respiratory distress syndrome (ARDS), intra‑abdominal hypertension, and infection in severely ill and burn patients. Over his career, he has authored numerous scientific publications, with hundreds of citations, and is now leading efforts to apply artificial intelligence and advanced monitoring to improve outcomes in critically ill patients