OncoReg: Medical Image Registration for Oncological Challenges
The OncoReg challenge benchmarks deformable image registration between low-dose CBCTs and high-resolution FBCTs for radiotherapy planning. Strongest methods consistently used robust feature extraction, such as ConvexAdam and DINO-Reg, proving that effective feature representation is central to modern registration pipelines. The work emphasizes the importance of preprocessing and handling partial image overlap for better results.