Spectral CT based radiomics for predicting brain metastases in patients with lung cancer

Background : The goal of this study was to create a prediction model for brain metastasis (BrMs) in patients with lung cancer using unenhanced spectral computed tomography (CT) and radiomics. Materials and Methods: This study comprised 162 patients with lung cancer who underwent spectral CT from 2019-2021. Patients were split into training and test sets and into BrMs and BrMs-free groups. Spectral and radiomics parameters were obtained from the spectral CT images before pathological confirmation. Prediction models in the training and test sets were created using logistic regression. The receiver operating characteristic curve was used to evaluate each quantitative parameter for predicting BrMs. The diagnostic effectiveness of several parameters was analyzed and compared using the area under the curve (AUC) calculation. The final model was obtained using the Delong test. Results: There were statistically significant differences in the iodine concentrations and the slope of the energy spectrum attenuation curve of the two groups <(p0.05). The AUC of the combined radiomics model was greater than that of the 70 keV and 120 keV sequence models. The joint parameters of radiomics and spectral CT constructed an integrated model. In the training set, test set, and overall set, the AUCs of the integrated model were 0.875, 0.879, and 0.724, respectively. In the training and overall sets, the prediction performance of the integrated model outperformed the spectral and radiomics models (p<0.05). Conclusions: This integrated model may predict the BrMs in lung cancer patients.