Artificial Intelligence Versus Anesthesiology Residents in Chest X-Ray Interpretation

Condition(s) studied

Study summary

This observational study will compare how accurately anesthesiology residents and an artificial intelligence system (hChestXR) identify abnormalities on chest X-rays. Approximately 35 volunteer anesthesiology residents at Kayseri City Hospital will review the same set of 150 anonymized posteroanterior chest radiographs selected from the hospital imaging archive. The set will include normal images and images showing findings such as infiltration, atelectasis, and pleural effusion. Two radiologists will independently assess the images and resolve disagreements by consensus to establish the reference standard. Residents will not see the radiologists' assessments or the artificial intelligence results while interpreting the images. The study will measure diagnostic accuracy and other measures of diagnostic performance. No additional imaging, treatment, or change in patient care is planned.

Eligibility

Primary outcome measure(s)

• Diagnostic Accuracy for Detection of Thoracic Abnormalities — During completion of the 150-radiograph assessment, within the 1-month study period
  Percentage of correctly classified image-finding pairs, calculated as (true positives + true negatives) / all evaluated image-finding classifications × 100, using the consensus of two radiologists as the reference standard. Accuracy will be calculated for each resident and for hChestXR overall and separately for each predefined thoracic finding.

Trial sites (1)

More Çiğdem Ünal Kantekin trials in Turkey

Other trials for the same condition