Performance validation of an artificial intelligence-assisted chest radiograph algorithm for incidental pulmonary nodule detection in Malaysian healthcare facilities: a multicentre cross-sectional study protocol.
Authors
Affiliations (3)
Affiliations (3)
- Public Health Medicine, Universiti Kebangsaan Malaysia, Cheras, Kuala Lumpur, Malaysia.
- Institut Kanser Negara, Putrajaya, Putrajaya, Malaysia.
- Public Health Medicine, Universiti Kebangsaan Malaysia, Cheras, Kuala Lumpur, Malaysia [email protected].
Abstract
Incidental pulmonary nodules (IPNs) are commonly encountered on chest radiographs (CXRs) performed for routine clinical indications and may represent early manifestations of significant pulmonary pathology, including lung cancer. While low-dose CT screening has mortality benefits in selected high-risk populations, its implementation remains limited in many healthcare settings. Artificial intelligence (AI)-assisted CXR interpretation has the potential to enhance pulmonary nodule detection. However, evidence from Malaysian clinical practice is scarce. This study aims to evaluate the diagnostic performance of AI-assisted CXR interpretation for detecting IPNs across healthcare facilities in the Klang Valley, Malaysia. This prospective, multicentre study will include 2452 CXRs from patients aged ≥35 years over a 6-month period across four Klang Valley healthcare facilities. Each CXR will be independently interpreted by an experienced radiologist (>5 years of experience) and analysed separately using an AI system (qXR-LNMS). An independent thoracic radiologist will determine the final classification for analysis if there is IPN detection discordance. Diagnostic performance metrics (sensitivity, specificity, positive and negative predictive values, and overall accuracy) will be calculated using a 2×2 classification matrix. Agreement between AI-assisted interpretation and radiologist reports will be assessed using Cohen's kappa statistic. The prevalence of IPNs detected by AI-assisted interpretation and radiologist reporting will be compared using a two-proportion z-test. AI discriminative performance will be evaluated using receiver operating characteristic curve analysis and area under the curve estimation. Statistical analyses will be conducted using Statistical Package for the Social Sciences V.29, with p<0.05 considered statistically significant. Ethical approval has been obtained from the Universiti Kebangsaan Malaysia Research Ethics Committee and the Ministry of Health Malaysia Medical Research and Ethics Committee. Written informed consent will be obtained from all participants. The findings will be disseminated through peer-reviewed publications, scientific conferences and engagement with relevant stakeholders.