A study in 70 patients with untreated non-small cell lung cancer (NSCLC) investigated the non-invasive prediction of stromal fibrosis (via α-SMA) and CD8+ T-cell density in the tumor microenvironment using dual-energy CT (DECT) and radiomics. Preoperative DECT scans provided DECT parameters and radiomic features, which were compared with analysis of surgical specimens to determine the percentage of α-SMA-positive stroma and CD8+ T-cell density; patients were divided into high and low level groups. Models for the prediction of high α-SMA expression showed the best performance of the integrated model (AUC 0.766), compared to DECT alone (AUC 0.670) and radiomics (AUC 0.703). For CD8+ T-cell density, the DECT model (AUC 0.715) was comparable to radiomics (AUC 0.695) without the advantage of integration. Key features included normalized iodine concentration for α-SMA and K40-70 spectral slope for CD8+ T-cells with significant differences between groups. Integration of DECT and radiomics is a feasible noninvasive method for TME evaluation in NSCLC.