Enhancing the prediction of KRAS mutation status in Asian lung adenocarcinoma: a comprehensive approach combining clinical, dual-energy spectral computed tomography, and radiomics features.

Abstract

Background: Lung adenocarcinoma (LUAD) is a sub-type of non-small cell lung cancer (NSCLC) that is often associated with genetic alterations, including the Kirsten rat sarcoma viral oncogene homolog (KRAS) mutation. The KRAS mutation is particularly challenging to treat due to resistance to targeted therapies. This study aims to develop a predictive model for the KRAS mutation in patients with LUAD by integrating clinical, dual-energy spectral computed tomography (DESCT), and radiomics features.

Methods: A total of 172 patients with LUAD were retrospectively enrolled and divided into a developing cohort (n=120) and a validation cohort (n=52). Clinical, DESCT and radiomics features were extracted and analyzed. Four predictive models were constructed: clinical, DESCT, radiomics, and combined clinical-DESCT-radiomics (C-S-R) model. The performance of these models was evaluated by the receiver operating characteristic curves. A nomogram incorporating clinical, DESCT, radiomics features with R-score was developed in the validation cohort.

Results: In this study, 8.7% (15/172) of the patients showed KRAS mutation. The C-S-R model demonstrated the best performance, with an area under the curve (AUC) of 0.92 in the developing cohort and 0.87 in the validation cohort. The C-S-R model was not superior to radiomics model (P=0.28), but it was significantly better than DESCT model (P=0.01).

Conclusions: This study suggests that integrating clinical, DESCT, and radiomics features can enhance the prediction of KRAS mutation in patients with LUAD.