Machine learning of CT-derived vascular morphology distinguishes symptomatic from asymptomatic renal vein compression

Nutcracker syndrome (NCS) and asymptomatic nutcracker phenomenon (ANP) share similar imaging features but require markedly distinct clinical management strategies. This study analyzed CT images from 60 patients (30 NCS and 30 ANP cases), employing 3D Slicer to extract geometric parameters of the left renal vein, including vascular angle, diameter, cross-sectional area, and volume, combined with patient age. A multilayer perceptron (MLP) classification model was constructed, achieving an AUC of 0.92 under five-fold cross-validation. Vascular angle and minimum diameter exhibited the highest feature weights in the model. The multilayer perceptron showed superior diagnostic performance to logistic regression, achieving an area under the curve of 0.92 compared to 0.83 for logistic regression. This study proposes an objective methodology for distinguishing NCS from ANP, though future validation with larger cohorts is warranted to assess generalizability.