Imaging in ARDS Moves Towards More Targeted Care

ARDS management is moving from broad syndrome-based treatment towards more targeted, physiology-guided care. Treatment has long relied on the idea that different causes of severe hypoxaemia share common biological pathways. That approach has supported standardised care and improved clinical outcomes, but it can also overlook major differences between patients. A 2026 Intensive Care Medicine update on imaging in ARDS places lung imaging at the centre of this shift. Current imaging methods, including conventional techniques, artificial intelligence tools and radiomics, can show how the injured lung changes in structure and function. These methods may help identify measurable treatment targets. The key challenge is to connect imaging information with practical decisions on ventilation, prone positioning, medicines and procedures.

From Standard Treatment to Individual Targets

Core ARDS treatment includes low-tidal volume ventilation, positive end-expiratory pressure (PEEP) and prone positioning in moderate-to-severe cases. Several treatments have looked promising in preclinical work or small human investigations, especially when measured through physiological or surrogate outcomes. Large randomised controlled trials have often failed to show a mortality benefit for treatments such as systemic corticosteroids, other immunomodulatory agents, inhaled nitric oxide, extracorporeal membrane oxygenation, statins, recruitment manoeuvres and uniform higher-PEEP strategies.

One recurring problem is poor physiological targeting. ARDS can be treated as one broad syndrome, while individual patients may differ greatly in lung damage, recruitability, inflammation and perfusion. Lung imaging has been part of the ARDS definition since the beginning, but its role now extends beyond confirming lung involvement. Conventional imaging and AI-based methods can show patterns that may help clinicians select treatments more precisely.

In 2019, the LIVE trial tested whether mechanical ventilation tailored to lung morphology could improve survival compared with standard care. The intention-to-treat result was neutral, mainly because clinicians misclassified lung morphology. Even so, the trial supports the idea that imaging can identify diffuse and focal, mainly dorsal, forms of ARDS. These forms may respond differently to higher PEEP and prone positioning. The main lesson is that imaging can help reveal treatment-relevant differences, but accurate classification remains essential.

Phenotyping Still Needs Practical Value

Efforts to divide ARDS into subtypes are not new. In the 1990s, a distinction between pulmonary and extrapulmonary ARDS did not become useful for clinical decision-making because it did not separate patients well enough and did not lead to clear treatment choices. More recent work has identified hypo-inflammatory and hyperinflammatory phenotypes from clinical trial data. Early interest increased when secondary analyses of previously neutral randomised trials suggested that simvastatin might benefit the hyperinflammatory subgroup.

Over the past decade, ARDS stratification has created many patient groupings through biomarkers, clinical characteristics and computational methods. However, these groupings have had limited impact on bedside care. Prospective validation has largely been missing, so precision medicine approaches based on these tools are still far from routine use. The problem is not only identifying different patient profiles. The bigger issue is linking each profile to a specific action that can improve care.

Must Read: Upright CT Clarifies Hidden Lung Regions

Recent data show that inflammatory subphenotypes can be identified quickly using a bedside test. That ability does not yet prove that treatment can be safely or effectively selected on that basis. Imaging may help close part of this gap because it captures rapid changes in ARDS, especially when combined with respiratory monitoring. Its value depends on moving from simple classification towards features that guide defined interventions at the right time.

Imaging, AI and Bedside Decisions

AI expands the role of imaging in ARDS through radiomics, which extracts quantitative information from medical images. Some of these features may not be visible during routine inspection but may reflect underlying biological processes. Foundation models and vision-language systems are now being applied to chest imaging. These tools can support automated segmentation, pattern recognition and the combination of imaging with clinical and molecular data.

Important limitations remain. Most AI-based imaging models in ARDS have been trained on single-centre retrospective data and have limited external validation. This creates concerns about generalisability, algorithmic bias, interpretability and excessive reliance on automated tools in clinical decisions. Even when imaging features can be extracted, the next step remains difficult. Imaging-derived information needs to guide ventilator settings, medicines or procedures, but that route from measurement to treatment is still largely unproven.

PEEP titration shows both the promise and the current gap. There is broad agreement that PEEP should be tailored to lung recruitability. However, no large randomised controlled trial has systematically tested PEEP titration based on quantitative computed tomography, electrical impedance tomography (EIT) or lung ultrasound. Recent progress includes EIT-guided PEEP setting associated with better respiratory mechanics and potentially better clinical outcomes. International consensus recommendations now also support standardised EIT-based bedside assessment of ventilation and perfusion. Commercial EIT-based perfusion assessment can quantify ventilation-perfusion mismatch at the bedside, and the extent of mismatch has been independently associated with mortality in ARDS.

Imaging has made ARDS more measurable, from chest radiography and lung ultrasound to computed tomography, EIT, dual-energy CT, radiomics and biomarker integration. Much of this information still has limited clinical use unless it connects to specific treatment targets. Lung inhomogeneity, damage distribution, ventilation-perfusion mismatch, inflammatory markers and multi-omics signals may help guide more targeted strategies. Combined with respiratory monitoring, gas exchange assessment and AI-driven analytics, imaging can support more practical ARDS management. Without a clear link to treatment, ARDS phenotypes remain theoretical rather than clinically useful.

Source: Intensive Care Medicine

Image Credit: iStock