Lightweight TotalSegmentator Integration for CT & MRI Anatomical Segmentation in InVesalius

This project tackles the challenge of limited anatomical segmentation in InVesalius by integrating TotalSegmentator through a lightweight inference pipeline using pre-exported JIT and ONNX models. The approach uses nnU-Net-based models to automatically segment 117 CT structures and 80 MRI structures: organs, bones, muscles, and vessels; without requiring nnunetv2 or any heavy dependencies. Users select specific anatomical structures from a categorized checklist and receive color-coded masks directly within InVesalius's existing visualization, editing, and 3D export pipeline. This enables clinicians in resource-constrained hospitals to perform comprehensive anatomical segmentation on the same workstations they already use.