Precise diagnosis of small invasive pulmonary nodules driven by single-cell immune signatures in peripheral blood.

Summary

In a multicenter prospective study, mass cytometry-derived peripheral immune signatures combined with machine learning distinguished invasive from non-invasive pulmonary nodules (AUC 0.952) and predicted adenocarcinoma invasiveness (AUC 0.949). The platform outperformed clinical and radiomics models, indicating immediate translational potential to guide surgery and reduce overtreatment.

Key Findings

• Peripheral immune signatures classified invasive versus non-invasive pulmonary nodules with AUC 0.952, surpassing clinical and radiomics models.
• Predicted tumor invasiveness, differentiating minimally invasive from invasive adenocarcinoma with AUC 0.949.
• Prospective, multicenter design supports translational generalizability for clinical decision support.

Clinical Implications

Could inform surgical versus surveillance decisions for small pulmonary nodules, reduce unnecessary resections, and prioritize high-risk lesions. Implementation would require standardized immune profiling and validated ML pipelines integrated into lung nodule clinics.

Limitations

• Sample size and external validation cohorts are not specified in the abstract
• Operational deployment requires access to mass cytometry and robust, generalizable ML pipelines

Future Directions

Conduct large-scale external validation, assess real-world impact on management and outcomes, standardize panels and analytics, and evaluate cost-effectiveness in lung nodule programs.