Machine learning-based multiparametric MRI radiomics nomogram for predicting WHO/ISUP nuclear grading of clear cell renal cell carcinoma.

Objective: To explore the effectiveness of a machine learning-based multiparametric MRI radiomics nomogram for predicting the WHO/ISUP nuclear grading of clear cell renal cell carcinoma (ccRCC) before surgery.

Methods: Data from 86 patients who underwent preoperative renal MRI scans (both plain and enhanced) and were confirmed to have ccRCC were retrospectively collected. Based on the 2016 WHO/ISUP grading standards, patients were divided into a low-grade group (Grade I and II) and a high-grade group (Grade III and IV), and randomly split into training and testing sets at a 7:3 ratio. Radiomics features were extracted from FS-T2WI, DWI, and CE-T1WI sequences. Optimal features were selected using the Mann-Whitney U test, Spearman correlation analysis, and the least absolute shrinkage and selection operator (LASSO). Five machine learning classifiers-logistic regression (LR), naive bayes (NB), k-nearest neighbors (KNN), adaptive boosting (AdaBoost), and multilayer perceptron (MLP)-were used to build models to predict ccRCC WHO/ISUP nuclear grading. The model with the highest area under the curve (AUC) in the testing set was chosen as the best radiomics model. Independent clinical risk factors were identified using univariate and multivariate logistic regression to create a clinical model, which was combined with radiomics score (rad-score) to develop a nomogram. The model's effectiveness was assessed using the receiver operating characteristic (ROC) curve, its calibration was evaluated using a calibration curve, and its clinical utility was analyzed using decision curve analysis.

Results: Six radiomics features were ultimately selected. The MLP classifier showed the highest diagnostic performance in the testing set (AUC=0.933). Corticomedullary enhancement level (P=0.020) and renal vein invasion (P=0.011) were identified as independent risk factors for predicting the WHO/ISUP nuclear classification and were included in the nomogram with the rad-score. The ROC curves indicated that the nomogram model had strong diagnostic performance, with AUC values of 0.964 in the training set and 0.933 in the testing set.

Conclusion: The machine learning-based multiparametric MRI radiomics nomogram provides a highly predictive, non-invasive tool for preoperative prediction of WHO/ISUP nuclear grading in patients with ccRCC.