Evaluation of the ‘qXR’ Software for the Detection of Pulmonary Nodules and Signs Suggestive of Heart Failure: A Comparative Analysis in a Latin American General Hospital

Background/Objectives: Artificial intelligence (AI) tools for chest X-ray interpretation have gained relevance as support systems in diagnostic workflows, particularly in settings with high demand or limited specialist availability. This study evaluated the diagnostic performance of the qXR software (Qure.ai) for detecting high-risk pulmonary nodules, cardiomegaly, and pleural effusion in adult patients at Hospital Clínica Bíblica in San José, Costa Rica. Methods: Three radiologists independently interpreted 225 chest radiographs, serving as the reference standard. qXR results were compared against this standard for each finding. Sensitivity, specificity, Cohen’s kappa, and area under the curve (AUC) were calculated. Predictive values were not used for interpretation due to the artificial prevalence of the sample. Results: qXR showed higher agreement with radiologist assessments for pulmonary nodules and pleural effusion, achieving moderate to substantial concordance. Performance for cardiomegaly was more variable, with lower agreement across evaluators. Overall diagnostic accuracy was acceptable, although the magnitude differed by condition. Conclusions: These findings underscore the importance of validating AI diagnostic tools within local clinical environments and heterogeneous imaging conditions. qXR demonstrated potential as a complementary aid for detecting pulmonary nodules and pleural effusion, while its performance for cardiomegaly should be interpreted with caution. The study does not provide evidence of real-world clinical impact.