Abstract
Lung cancer (LC) remains a leading global cause of cancer mortality, with current diagnostic and prognostic methods lacking precision. This meta-analysis evaluated the role of artificial intelligence (AI) in LC imaging-based diagnosis and prognostic prediction. We systematically reviewed 315 studies from major databases up to January 7, 2025. Among them, 209 studies on LC diagnosis yielded a combined sensitivity of 0.86 (0.84-0.87), specificity of 0.86 (0.84-0.87), and AUC of 0.92 (0.90-0.94). For LC prognosis, 106 studies were analyzed: 58 with diagnostic data showed a pooled sensitivity of 0.83 (0.81-0.86), specificity of 0.83 (0.80-0.86), and AUC of 0.90 (0.87-0.92). Additionally, 53 studies differentiated between low- and high-risk patients, with a pooled hazard ratio of 2.53 (2.22-2.89) for overall survival and 2.80 (2.42-3.23) for progression-free survival. Subgroup analyses revealed an acceptable performance. AI exhibits strong potential for LC management but requires prospective multicenter validation to address clinical implementation challenges.