Automated MRI-Based Segmentation of the Cranio-Cervical Venous System Using a Dedicated Deep Learning Pipeline

PURPOSE To develop an automated magnetic resonance imaging based segmentation pipeline for the cranio-cervical venous system (CCVS), a complex anatomical network for which no dedicated segmentation tool is currently available. MATERIALS AND METHODS A retrospective single-center cohort of 194 contrast-enhanced cranio-cervical MRI examinations acquired between 2019 and 2024 was used. Manual reference segmentations were performed on 50 examinations using 3D Slicer. A nnU-Net v2-based segmentation model was trained through an iterative process combining automatic segmentation, manual correction, and retraining. Segmentation performance was assessed using the Dice Similarity Coefficient (DSC). The complete processing pipeline was integrated into an open-source software application, VenousNetworkAnalyzer (VNA). RESULTS The final model achieved a mean DSC of 0.94 (95% CI, 0.93-0.95) in five-fold cross-validation on 106 examinations with manual reference segmentations. Iterative optimization progressively reduced manual processing time from approximately 10 hours of fully manual segmentation to less than 10 minutes of manual correction per patient. Application of the final model to the entire cohort enabled the generation of patient-specific three-dimensional reconstructions of the cranio-cervical venous system. The segmentation workflow was successfully integrated into the VNA software environment, allowing automated processing and visualization of MRI datasets. CONCLUSION This study demonstrates the feasibility of automated segmentation of the cranio-cervical venous system from contrast-enhanced MRI using a dedicated deep-learning pipeline. The resulting reduction in processing time and integration into an open-source software platform provide a foundation for future large-scale anatomical studies of cranio-cervical venous draina