In this retrospective analysis, the contrast-enhanced CT scans of 300 patients with confirmed non-small cell lung cancer, including 200 with adenocarcinoma and 100 with squamous cell carcinoma were assessed to evaluate the performance of radiomics as a virtual biopsy tool for histological classification. All scans were assessed for radiological features such as affected lung lobe, central/peripheral location within lobe, emphysema, and T/N tumor stage. Additionally, scans underwent three-dimensional tumor segmentation, allowing extraction of 107 radiomic features. The radiomics-based model outperformed the radiological model alone, achieving an area under the curve (AUC) of 0.80 compared to 0.73. The highest diagnostic performance was observed in the combined radiological-radiomics model, which integrates the two approaches, obtaining an AUC of 0.84 and an accuracy of 75.3%. These findings support the superiority of the combined model for an accurate and non-invasive differentiation between lung adenocarcinoma and squamous cell carcinoma, providing a promising virtual biopsy tool to predict the histological nature of non-small cell lung cancer.