Purpose: To establish a CT-based radiomics nomogram for preoperative prediction of KRAS mutation and prognostic stratification in colorectal cancer (CRC) patients.
Methods: In a retrospective analysis, 408 patients with confirmed CRC were included, comprising 168 cases in the training set, 111 cases in the internal validation set, and 129 cases in the external validation set. Radiomics features extracted from the primary tumors were meticulously screened to identify those closely associated with KRAS mutation. Subsequently, a radiomics nomogram was constructed by integrating these radiomics features with clinically significant parameters. The diagnostic performance was assessed through the area under the receiver operating characteristic curve (AUC). Lastly, the prognostic significance of the nomogram was explored, and Kaplan-Meier analysis was employed to depict survival curves for the high-risk and low-risk groups.
Results: A radiomics model was constructed using 19 radiomics features significantly associated with KRAS mutation. Furthermore, a nomogram was developed by integrating these radiomics features with two clinically significant parameters (age, tumor location). The nomogram achieved AUCs of 0.834, 0.813, and 0.811 in the training set, internal validation set, and external validation set, respectively. Additionally, the nomogram effectively stratified patients into high-risk (KRAS mutation) and low-risk (KRAS wild-type) groups, demonstrating a significant difference in overall survival (P < 0.001). Patients categorized in the high-risk group exhibited inferior overall survival in contrast to those classified in the low-risk group.
Conclusions: The CT-based radiomics nomogram demonstrates the capability to effectively predict KRAS mutation in CRC patients and stratify their prognosis preoperatively.
Keywords: Colorectal cancer; Computed tomography; KRAS mutation; Nomogram; Radiomics.
© 2024. The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.