Application of dual-energy computed tomography combined with radiomics in the clinical diagnosis of lung cancer: a systematic review and meta-analysis

Background: Lung cancer is the leading cause of cancer-related mortality globally, with early precise diagnosis critical for improving prognosis. Conventional computed tomography (CT) lacks sufficient sensitivity for early malignant nodule detection, while pathological biopsy is invasive and limited by sampling bias-creating an urgent need for non-invasive, high-accuracy diagnostic tools. This study aimed to systematically evaluate the diagnostic efficacy of dual-energy computed tomography (DECT) across different clinical scenarios of lung cancer and further explore the additional diagnostic value of its integration with radiomics models, aiming to provide an evidence-based reference for precise imaging assessment and clinical decision-making in lung cancer. Methods: We systematically searched relevant databases, screened eligible studies, applied quality assessment tools for comprehensive bias evaluation, and analyzed their diagnostic accuracy. Statistical analysis software was used to perform heterogeneity test and subgroup analysis and sensitivity analysis, while Spearman rank correlation test was employed to examine the threshold effect. Results: A total of 2,899 lesions were included across the 21 studies [2016-2025]. Differentiating benign from malignant: DECT achieved a pooled sensitivity of 0.87 [95% confidence interval (CI): 0.83-0.91], a specificity of 0.88 (95% CI: 0.79-0.93), and an area under the curve (AUC) of 0.93 (95% CI: 0.90-0.95). Integrated DECT-radiomics model showed training set a sensitivity of 0.90, a specificity of 0.88, an AUC of 0.94 (95% CIs: 0.85-0.93, 0.80-0.94, 0.85-1.00). Invasiveness prediction: DECT demonstrated a sensitivity of 0.83 (0.80-0.85), a specificity of 0.79 (0.76-0.83), and an AUC of 0.85 (0.81-0.88). DECT-radiomics integration improved performance, with training set a sensitivity of 0.84, a specificity of 0.84, and an AUC of 0.91 (95% CIs: 0.80-0.87, 0.79-0.87, 0.88-0.93). In addition, the diagnostic efficacy of DECT for predicting lymph node metastasis (LNM) was evaluated independently: a sensitivity of 0.84 (0.75-0.90), a specificity of 0.83 (0.75-0.89), an AUC of 0.90 (0.87-0.93). Subgroup regression analysis indicated that feature extraction methods and radiomics algorithms are the most important sources of heterogeneity. This study did not find significant threshold effect or publication bias (P>0.05). Conclusions: DECT demonstrates high diagnostic accuracy in assessing key characteristics of lung cancer. When integrated with radiomics, it significantly improves the performance of distinguishing benign from malignant lesions and enhances the accuracy of invasiveness prediction, thereby offering robust technical support for optimizing precision imaging assessment strategies in lung cancer. In future research, the clinical generalizability and translational potential of this combined technique could be further validated through an expanded sample size and the inclusion of lung cancer cases representing diverse pathological subtypes.