Risk-Routed Implicit Boundary Refinement for Robust Ultrasound Image Segmentation
The Risk-Routed Implicit Boundary Refinement (RIBR) framework refines ultrasound segmentation by using implicit neural representations as a risk-routed residual correction instead of a full predictor. It combines boundary refinement, risk routing, and geometry-aware regularization to produce compact models that achieve superior boundary accuracy across nine medical datasets.