Screening Landscape: Despite LDCT's proven mortality benefit, lung cancer screening reaches only a fraction of eligible individuals worldwide. New technologies are needed to improve detection accuracy, accessibility, and cost-effectiveness.
Review Scope: This systematic review covers PubMed literature up to May 2025, encompassing conventional imaging (CXR, LDCT), advanced bronchoscopy, and liquid biopsy modalities (CTCs, cfDNA, exosomes, methylation biomarkers) alongside AI-driven approaches.
Three Proposed Directions: The review proposes three development pathways: (1) risk-stratified liquid biopsy for population pre-screening, (2) an 'imaging-AI-liquid biopsy' integrated workflow, and (3) standardized biomarker panels for clinical translation.
Clinical Relevance: Early-stage (Stage I/II) lung cancer has a 5-year survival of 50-80%, versus below 10% for Stage IV, making effective early detection strategies one of the highest-impact interventions in oncology.
Chest X-Ray Limitations: While universally available and inexpensive, chest X-ray (CXR) is insufficient for lung cancer screening due to low sensitivity for small nodules, particularly those overlapping with ribs, vessels, or the mediastinum.
LDCT Evidence Base: The NLST and NELSON trials established LDCT screening in high-risk current and former smokers reduces lung cancer mortality by 20-26%. LDCT is now recommended by major guidelines, though implementation lags in many regions.
LDCT Limitations: High false-positive rates (23% in NLST), radiation exposure, incidental findings, and cost create practical barriers. High-risk nodules (4-8%) lead to invasive follow-up, most of which prove benign.
AI-Enhanced LDCT: Deep learning algorithms for automated nodule detection, malignancy risk scoring, and incidental finding triage are improving LDCT efficiency and reducing radiologist workload, with several systems achieving FDA clearance.
Navigational Bronchoscopy: Electromagnetic navigational bronchoscopy (ENB) and robotic bronchoscopy allow guided biopsy of peripheral lung nodules inaccessible to standard bronchoscopy, reducing the need for CT-guided percutaneous biopsy and its complication risks.
Autofluorescence and Narrow-Band Imaging: These endoscopic imaging modalities enhance visualization of pre-malignant central airway lesions that appear normal under white light, improving detection of central carcinoma in situ and early squamous cell lesions.
Cytological Yield: Bronchoalveolar lavage combined with advanced AI-based cytology analysis - as developed in studies like LESSEL - can detect malignant cells in BAL fluid from peripheral tumors, extending bronchoscopy's reach beyond the central airways.
Bronchial Thermoplasty and Biopsy Platforms: Combined bronchoscopic platforms integrating navigation, ultrasound, and biopsy capability are improving tissue yield from small peripheral lesions, reducing the sampling failure rate that previously limited bronchoscopy for small nodules.
Circulating Tumor Cells (CTCs): CTCs shed from primary or metastatic tumors into blood can be captured and analyzed for cancer-specific genomic or proteomic features, providing prognostic information and treatment monitoring capability, though sensitivity in early-stage disease is low.
Cell-Free DNA (cfDNA) and ctDNA: cfDNA from tumor cell apoptosis or necrosis carries cancer-specific mutations, copy number alterations, and methylation patterns that can detect lung cancer, monitor minimal residual disease, and identify resistance mutations.
Exosomes: Tumor-derived exosomes in blood carry RNA, proteins, and DNA from the originating cancer cells, providing a protected sampling of tumor molecular content. Exosome-based biomarkers show promise for early detection but require standardized isolation protocols.
DNA Methylation: Cancer-specific hypermethylation of CpG islands in plasma cfDNA enables detection of lung cancer with high specificity. Multi-cancer early detection (MCED) tests utilizing methylation patterns have demonstrated ability to detect lung cancer even at Stage I/II.
Nodule Detection AI: Deep learning models for automated pulmonary nodule detection on LDCT match or exceed radiologist sensitivity while dramatically reducing false-positive rates, enabling consistent performance across volume and scanner variability.
Malignancy Risk Prediction: AI-based malignancy risk scores (such as LCP-CNN) provide continuous probability estimates for individual nodules, complementing categorical Lung-RADS assignments and improving the clinical discrimination of borderline cases.
Multi-Omics AI Integration: Machine learning models combining imaging features with liquid biopsy data, genomic profiles, and clinical risk factors demonstrate superior performance compared to any single modality, though multicenter validation remains limited.
AI Workflow Optimization: Beyond detection and classification, AI assists with screening eligibility determination, follow-up interval recommendation, report generation, and identifying missed nodules in prior scans, improving end-to-end screening program efficiency.
Direction 1 - Risk-Based Liquid Biopsy Pre-Screening: Deploy blood-based biomarker panels (ctDNA, methylation, proteomics) as a first-tier risk stratification tool to identify the highest-risk individuals who should proceed to LDCT, reducing radiation and cost burden.
Direction 2 - Imaging-AI-Liquid Biopsy Workflow: Create an integrated diagnostic pathway where LDCT findings trigger AI-based risk assessment, which then triggers targeted liquid biopsy testing for borderline nodules, combining imaging and molecular information in a sequential decision tree.
Direction 3 - Standardized Biomarker Panels: Develop consensus, standardized multi-biomarker panels (validated in large prospective studies) that can be consistently applied across institutions, enabling reproducible, guideline-supported clinical use.
Implementation Priorities: Each direction requires regulatory validation, health economic analysis, and integration into existing clinical workflows - necessitating coordinated investment from governments, health systems, and industry.
Standardization Gap: Most liquid biopsy and AI technologies lack standardized pre-analytical and analytical protocols, making cross-study comparison difficult and delaying clinical implementation.
Early-Stage Detection Sensitivity: Liquid biopsy sensitivity for Stage I lung cancer remains 30-50% for most platforms, insufficient as a standalone screening test; further technological development or combination strategies are needed.
Population Diversity: Most screening and biomarker studies have been conducted in high-income countries with predominantly white or East Asian cohorts. Global representation in validation studies is needed for equitable global impact.
Prospective Randomized Evidence: The ultimate evidence standard - randomized controlled trials demonstrating mortality reduction from AI or liquid biopsy-augmented screening compared to LDCT alone - remains largely unavailable and is urgently needed.