A clinical-radiomics Nomogram based on spectral CT quantitative parameters for preoperative prediction of tumor budding grading in colorectal adenocarcinoma

To develop a clinical-radiomics Nomogram based on spectral CT quantitative parameters for non-invasive preoperative prediction of tumor budding(TB) grading in colorectal adenocarcinoma(CRAC). This retrospective study enrolled 138 patients with CRAC from our center between May 2021 and August 2025. Spectral parameter values were measured for each case. Univariate analysis was used to compare the difference of traditional and spectral parameters between different grades of TB. The Synthetic Minority Oversampling Technique (SMOTE) was applied to mitigate class imbalance, and multivariate dimensionality reduction including Pearson Correlation Coefficient (PCC), ANOVA, Relief algorithm, Recursive Feature Elimination (RFE) and Kruskal‑Wallis test was performed for feature screening. A sequential dimensionality reduction pipeline was applied to reduce 107 radiomics features to 23 candidate features for exploratory model construction. Radiomics models were constructed by extracting three‑dimensional volume of interest (VOI) features from spectral quantitative parameter images. Combining radiomics score, independent clinical indicator and spectral parameters, a clinical-radiomics Nomogram was constructed. The model with the highest area under the curve(AUC) value was selected for subsequent analysis, and its classification threshold was defined for reference in clinical exploration Model performance was assessed using AUC, calibration curves and decision curve analysis (DCA). Low-density lipoprotein cholesterol(LDL-C) and 40keV virtual monoenergetic image(VMI) in arterial phase were identified as independent predictors for TB grading. After SMOTE balancing and multivariate feature optimization, the combined clinical‑radiomics Nomogram exhibited relatively favorable predictive performance in internal validation, with training set AUC of 0.895[95% confidence interval(CI): 0.818‑0.972] and test set AUC of 0.781(95%CI: 0.688‑0.934). The optimal classification threshold of the Nomogram was determined to effectively distinguish low‑grade TB(Bd1) from moderate‑to‑high‑grade TB(Bd2+3). Calibration curves demonstrated good consistency between predicted and actual outcomes (all P>0.05). DCA indicated that the Nomogram possessed higher clinical net benefit than clinical model and radiomics model. This study developed a clinical-radiomics Nomogram based on spectral CT quantitative parameter images, serving as an effective tool for preoperative non-invasive personalized prediction of TB grading in CRAC.