ID #166 CAVS-NF1: Web-Based AI Tool for Longitudinal Volumetric Analysis of Anterior Visual Pathway in NF1

Background Optic pathway gliomas associated with neurofibromatosis type 1 (NF1-OPG) affect the anterior visual pathway (AVP), including the optic nerves, chiasm, and tracts. To enhance clinical applicability and accessibility of AVP morphology on MRI, we introduce CAVS-NF1 (Central AI-enabled Volumetric Service for NF1), a free, AI-powered, web tool for automated segmentation and volumetric analysis of the AVP from brain MRIs. CAVS-NF1 allows longitudinal analysis from multiple MRIs from the same patient. Methods We developed an automated pipeline utilizing a vision transformer-based AI segmentation model based on SwinUNETR. The model was trained on T1-weighted MRIs of 135 children with NF1-OPG from Children’s National Hospital (n = 60 from GE scanners) and Children’s Hospital of Philadelphia (n = 75 from Siemens scanners). Expert delineations of AVP were used for training. Post-segmentation, a template-based registration was employed to isolate the nerves and chiasm. For longitudinal analysis, we aligned the temporal volumes to compute relative changes in substructure volumes. The CAVS-NF1 framework does not require local installation and is accessible via a secure, HIPAA-compliant web interface. Results The AVP segmentation method achieved a volumetric overlap (Dice score) of 79.1 ± 8.2 %, outperforming the inter-observer variability in manual annotations (Dice: 75.6±6.2 %). The average runtime for a single analysis was 98 ± 3 seconds. In the longitudinal analysis workflow, volume change computations for AVP, nerves, and chiasm were successfully demonstrated with consistent alignment and robust segmentation performance. The CAVS-NF1 web platform is free, HIPAA-compliant, and available at https://nf1.hope4kids.io. Conclusion CAVS-NF1 enables fast, automated, and accurate longitudinal volumetric analysis of the AVP in NF1-OPG patients. The platform supports longitudinal analysis with side-by-side comparison of local changes in tumor volume over time. This AI-driven web tool enhances OPG monitoring with the potential of assessing response to therapy and is accessible to clinicians and researchers without requiring advanced technical infrastructure.