The potential value of dual-energy CT radiomics in evaluating CD8+, CD163+ and αSMA+ cells in the tumor microenvironment of clear cell renal cell carcinoma.

Abstract

This study aims to develop radiomics models and a nomogram based on machine learning techniques, preoperative dual-energy computed tomography (DECT) images, clinical and pathological characteristics, to explore the tumor microenvironment (TME) of clear cell renal cell carcinoma (ccRCC). We retrospectively recruited of 87 patients diagnosed with ccRCC through pathological confirmation from Center I (training set, n = 69; validation set, n = 18), and collected their DECT images and clinical information. Feature selection was conducted using variance threshold, SelectKBest, and the least absolute shrinkage and selection operator (LASSO). Radiomics models were then established using 14 classifiers to predict TME cells. Subsequently, we selected the most predictive radiomics features to calculate the radiomics score (Radscore). A combined model was constructed through multivariate logistic regression analysis combining the Radscore and relevant clinical characteristics, and presented in the form of a nomogram. Additionally, 17 patients were recruited from Center II as an external validation cohort for the nomogram. The performance of the models was assessed using methods such as the area under the receiver operating characteristic curve (AUC), calibration curve, and decision curve analysis (DCA). The validation set AUC values for the radiomics models assessing CD8+, CD163+, and αSMA+ cells were 0.875, 0.889, and 0.864, respectively. Additionally, the external validation cohort AUC value for the nomogram reaches 0.849 and shows good calibration. Radiomics models could allow for non-invasive assessment of TME cells from DECT images in ccRCC patients, promising to enhance our understanding and management of the tumor.