Key points are not available for this paper at this time.
Acute lung injury and acute respiratory distress syndrome (ARDS) remain among the most heterogeneous and clinically challenging syndromes in critical care. Although lung-protective ventilation, prone positioning, fluid optimization, neuromuscular blockade, and extracorporeal support have improved outcomes, mortality and long-term morbidity remain substantial. A major reason is that ARDS is increasingly recognized as a biologically and physiologically heterogeneous syndrome rather than a single disease entity. Artificial intelligence (AI) is beginning to address this gap. Beyond outcome prediction, AI is now being applied to early recognition, automated case identification, chest imaging interpretation, lung ultrasound and electrical impedance tomography enhancement, patient–ventilator asynchrony detection, phenotyping, treatment response prediction, and closed-loop support in intensive care. The new global definition of ARDS further broadens the relevance of AI by extending recognition to earlier-stage patients, to those receiving high-flow oxygen, and to resource-limited settings. The principal value of AI in lung injury lies not in replacing clinicians but in integrating multimodal and longitudinal data into a more continuous and computable representation of disease. When coupled with interpretable modelling, multicenter standardized datasets, and prospective validation, AI could help move lung injury care beyond syndrome-based supportive treatment toward a more individualized, dynamic, and computationally informed model of precision medicine.
Liang et al. (Mon,) studied this question.