Lung disease kills millions annually, yet diagnosis still depends heavily on overburdened clinicians interpreting complex, heterogeneous data across imaging, genomics, and clinical history. A comprehensive review now maps exactly how artificial intelligence is restructuring that entire workflow — and where the real translational barriers lie.
The synthesis covers AI performance across the full respiratory care continuum for conditions including lung cancer, COPD, asthma, interstitial lung diseases, and cystic fibrosis. On the diagnostic end, convolutional and transformer-based models have reached radiologist-level accuracy in pulmonary nodule detection and automated pattern classification. The review then traces the progression toward AI-powered clinical decision support systems embedded in real care workflows, as well as AI-guided surgical and radiotherapeutic planning. Perhaps most forward-looking is its treatment of "digital twin" simulations — patient-specific computational models that allow longitudinal disease trajectory modeling without additional biological sampling. The authors also assess how multi-omics integration and liquid biopsy analysis are being paired with machine learning to surface novel biomarkers and accelerate drug repurposing pipelines. Large language models and multimodal foundation models receive dedicated attention as the emerging frontier, distinct in capability from classical machine learning architectures.
This review arrives at a critical inflection point: the technology is demonstrably capable, yet clinical deployment remains fragmented. The authors' identification of sociotechnical barriers — data sovereignty, legal liability, and the near-universal absence of prospective validation — is where this synthesis earns its analytical weight. Most published AI respiratory studies are retrospective, single-institution, and lack external validation cohorts, meaning reported performance metrics routinely overstate real-world utility. The proposed translational roadmap acknowledges this gap explicitly. For health-conscious adults, the practical implication is that AI-assisted lung screening and personalized COPD management are approaching clinical readiness, but equitable deployment across diverse populations remains the field's most underresolved challenge.