Predicting NRAS gene status in colorectal cancer using computed tomography (CT)-based radiomics
1Department of Radiology, Yunnan Cancer Hospital, The Third Affiliated Hospital of Kunming Medical University, Peking University Cancer Hospital Yunan, Kunming, 650118, China.
Aim:
Establish and validate a radiomics models based on computed tomography (CT) images to predict NRAS status in colorectal cancer patients.
Materials And Methods:
Clinical and imaging characteristics associated with NRAS status were analysed. Imaging data from 216 colorectal cancer patients at hospital A were used, with 151 patients assigned to the training set and 65 to the test set. Regions of interest (ROIs) were delineated on enhanced venous phase CT images using ITK-SNAP. Features were selected based on analysis of variance, with logistic regression as the classifier. Five-fold cross-validation was applied to ensure model stability and generalisability. The Hosmer-Lemeshow test was used to evaluate the calibration curve, while clinical decision curve analysis assessed the clinical utility of the model.
Results:
No significant association was found between clinical features and NRAS gene status (P < 0.05). Five radiomic features were selected for model construction. The model achieved an area under the curve (AUC) of 0.803 (95% confidence interval [CI]: 0.750-0.864) in the training set and 0.766 (95% CI: 0.730-0.876) in the test set. Calibration curves demonstrated excellent agreement between the predicted and actual probabilities in both the training and test sets, supported by nonstatistically significant differences in the Hosmer-Lemeshow test (χ2 = 6.727, P = 0.56; χ2 = 4.854, P = 0.77). Clinical decision curve analysis demonstrated the clinical value of radiomics in predicting NRAS status.
Conclusion:
The CT-based radiomics model effectively predicts NRAS gene status in colorectal cancer patients.
