Artificial Intelligence (AI), Intensive Care Unit (ICU), Pneumonia, Bacterial, Radiography, Thoracic, Ventilator-Associated Pneumonia
Conditions
Brief summary
Ventilator-associated pneumonia (VAP) is a common and serious infection in critically ill patients receiving mechanical ventilation in intensive care units (ICUs). One of the key diagnostic criteria for VAP is the presence of a new or progressive infiltrate on chest X-ray; however, interpretation of bedside chest radiographs is often challenging and subject to inter-observer variability. This retrospective observational study aims to evaluate the role of artificial intelligence (AI) in the assessment of chest X-rays in patients with VAP. Chest radiographs obtained before and at the time of VAP diagnosis will be analyzed using a deep learning-based AI tool (Chester the AI Radiology Assistant), and changes in "infiltration" and "pneumonia" probability scores will be assessed. AI-based findings will be compared with clinical decisions and independent radiologist evaluations regarding the presence of new infiltrates. The study aims to determine the level of agreement between these approaches and to explore whether AI-based analysis can support a more objective and standardized interpretation of chest radiographs in the diagnosis of VAP.
Interventions
None listed
Sponsors
Dr Abdurrahman Yurtaslan Ankara Oncology Training and Research Hospital
Study design
Observational model
COHORT
Time perspective
RETROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
18 Years to No maximum
Healthy volunteers
No
Inclusion criteria
* Adult patients (≥18 years) * Admission to the anesthesia intensive care unit * Requirement of invasive mechanical ventilation for at least 48 hours * Clinical diagnosis of ventilator-associated pneumonia (VAP) based on institutional criteria * Availability of at least one chest X-ray prior to VAP diagnosis and one chest X-ray at the time of diagnosis * Availability of digital chest radiographs in the PACS system suitable for analysis
Exclusion criteria
* Age \<18 years * Absence of accessible digital chest radiographs * Chest radiographs with severe technical limitations preventing evaluation * Chest radiographs that could not be processed by the AI system (no score generated) * Presence of extensive pre-existing infiltrative lung disease preventing reliable assessment of new infiltrates * Missing key clinical data (e.g., VAP diagnosis date, mechanical ventilation duration)
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Change in Chester AI-derived Infiltration and Pneumonia Probability Scores Between Pre-diagnosis and VAP Diagnosis Chest X-rays | From pre-diagnosis chest X-ray to the time of VAP diagnosis (typically within 1-5 days) | The primary outcome is the change in probability scores for "infiltration" and "pneumonia" generated by the Chester AI Radiology Assistant between chest X-rays obtained prior to VAP diagnosis and those obtained at the time of diagnosis. These scores range from 0 to 1 and represent the likelihood of the presence of each finding. |
Countries
Turkey (Türkiye)
Outcome results
None listed