Radiomics in lung cancer: advances and challenges for clinical practice

Fernando A. Ramos-Zaga, Dirección de Investigación, Universidad Privada del Norte, Lima, Perú

Artificial intelligence–based radiomics has emerged as an approach capable of transforming medical imaging into quantitative biomarkers, in a context where lung cancer remains the leading cause of cancer-related mortality worldwide. The objective of this article is to analyze the current state of this technology as a non-invasive phenotypic characterization tool, assessing its ability to extract biomarkers and the standardization challenges that condition its clinical implementation. The findings indicate that radiomic models can predict relevant molecular alterations, estimate response to immunotherapy, and improve prognostic stratification with clinically meaningful performance; however, limitations persist related to variability in image acquisition, tumor segmentation, reproducibility, and model interpretability. It is concluded that, although its potential is substantial, the clinical adoption of radiomics requires robust standardization, multicenter validation, and governance frameworks to ensure its safe integration into precision medicine.

Keywords: Quantitative imaging biomarkers. Deep learning. Clinical translation.