The combined model, incorporating radiomic features from both traditional and habitat analysis along with clinical variables, outperformed the traditional radiomics-only and habitat-only models. The combined model achieved an area under the curve (AUC) of 0.862 in the training set and 0.843 in the validation set, demonstrating high predictive value for identifying PNI preoperatively. Additionally, the combined model exhibited superior sensitivity and specificity compared to the individual models, highlighting its potential clinical utility.