Radiomics in lung cancer surgery: from experimental promise to clinical translation—a critical narrative review

Background and Objective: Radiomics enables the extraction of quantitative imaging features that reflect tumour phenotype and intratumoral heterogeneity, offering a non-invasive approach to precision medicine. In thoracic surgery, radiomics has emerged as a promising tool to improve risk stratification and guide treatment strategies. This narrative review summarises the current evidence on radiomics applications in lung cancer across the thoracic surgical pathway and evaluates its potential integration into clinical practice.

Methods: A literature search was conducted using PubMed/MEDLINE, Scopus, and Google Scholar for studies published up to January 2026, combining Medical Subject Headings (MeSH) and free-text terms related to lung cancer and radiomics with surgical keywords. Original articles, systematic reviews, and prospective or retrospective cohort studies were considered. Article selection was performed through independent screening followed by full-text evaluation when necessary, with discrepancies resolved by consensus. The retrieved literature was organised thematically according to the stages of the thoracic surgical pathway.

Key Content and Findings: Radiomics has demonstrated utility across multiple clinically relevant domains. Radiomics-based models can improve pulmonary nodule characterisation and aid in distinguishing invasive from pre-invasive lesions. Radiomics signatures have shown promising performance in predicting lymph node metastases, occult nodal disease, and aggressive histologic patterns, including micropapillary and solid predominant components, spread through air spaces (STAS), and lymphovascular invasion (LVI), which may influence the extent of surgical resection. Radiomics features have also been associated with postoperative complications, recurrence risk, long-term survival, and adjuvant therapy benefit. Across these applications, radiomics models frequently demonstrate incremental value over conventional parameters. However, most studies remain retrospective with heterogeneous methodologies and limited external validation, and the predominance of positive results in the published literature likely reflects publication bias.

Conclusions: Radiomics represents a promising adjunct in lung cancer surgery. Its cautious incorporation as a decision-support tool appears feasible in specialised centres with standardised imaging protocols, provided that model outputs are interpreted within multidisciplinary frameworks and not as independent determinants of surgical strategy. Wider implementation will require prospective multicentric validation, methodological standardisation, and integration with clinical, functional, and molecular data.