Integrative Radio-Genomic Analysis of Lung Adenocarcinoma: Linking Imaging Features and Gene Expression Profiles for Tumor Classification

Lung adenocarcinoma (LUAD) diagnosis has traditionally been supported by both pathological imaging and molecular profiling; however, direct comparisons of these modalities for disease characterization remain limited. In this study, an iterative radiogenomic pipeline was implemented to investigate the correspondence between imaging-derived features and transcriptomic activity. Radiomic features were extracted from pathological images, while transcriptomic profiles were obtained from RNA sequencing (RNA-seq) data. Both modalities were analyzed independently to assess their ability to classify LUAD cases as early stage (Stage I) or advanced stage. Classification based on radiomic features alone achieved an accuracy of 85%, whereas RNA-seq data provided 100% accuracy.Feature selection was performed separately for each dataset, yielding highly discriminative radiomic features and significantly altered genes associated with LUAD staging. Correlation analysis was subsequently undertaken to explore cross-modality associations. Strong relationships were observed between specific imaging features and gene expression profiles, enabling the identification of genes most strongly linked to radiomic traits. These results suggest that radiomic signatures may, in some contexts, act as surrogates for transcriptomic activity, thereby reflecting underlying molecular mechanisms.The emphasis of this work was not placed on integrating imaging and molecular data to maximize classification accuracy, but rather on uncovering biological links between tumor morphology and gene expression. This correlation-driven framework demonstrates that molecular information may eventually be inferred from imaging alone, providing a cost-effective alternative to RNA-seq–based profiling. The findings highlight the potential of radiogenomic correlations to reveal biomarkers, support LUAD staging, and inform precision oncology strategies.