Key points are not available for this paper at this time.
Lung cancer remains the leading cause of cancer-related mortality worldwide, primarily due to late-stage diagnosis. Early detection is critical for improving survival outcomes, yet current screening modalities, particularly low-dose computed tomography, are limited by high false-positive rates, overdiagnosis concerns, and suboptimal population coverage. Recent advances in artificial intelligence (AI) and blood-based liquid biopsy technologies have opened new avenues for non-invasive, accurate, and cost-effective lung cancer early screening. This narrative review systematically examines the convergence of AI-driven analytical frameworks with two complementary data modalities: blood-based molecular biomarkers (including circulating tumor DNA, DNA methylation signatures, microRNAs, protein panels, circulating tumor cells, and exosomal cargo) and pathological/radiological imaging (encompassing histopathological whole-slide images, cytopathology, low-dose CT radiomics, and PET-CT features). We critically evaluate the diagnostic performance, clinical validation status, and translational potential of each biomarker category and imaging modality. Furthermore, we discuss the current landscape of AI and machine learning algorithms—from convolutional neural networks and vision transformers to multi-modal fusion architectures—that have been deployed to analyze these heterogeneous data streams for lung cancer detection, classification, and risk stratification. We also address major challenges, including data standardization, model interpretability, prospective validation, regulatory considerations, and health equity implications. Finally, we outline future directions for integrating multi-modal AI screening platforms into routine clinical practice to achieve population-level early lung cancer detection. By synthesizing evidence from over 185 key publications, we provide a comprehensive roadmap for researchers and clinicians seeking to harness these converging technologies for improved lung cancer outcomes.
Ling et al. (Mon,) studied this question.