Registration is a Representation Problem: Solving Whole-Body PET/CT Registration
We present an approach that instead establishes correspondence using complementary representations of anatomy and demonstrate its utility for longitudinal whole-body PET/CT registration in the Learn2Reg 2026 PSMAReg Challenge.
ProofPaper ↗
Key points
- Raw image intensities are a poor surrogate for similarity across images in correspondence problems.
- Our 3D foundation model, anatomix, is used to derive fine-grained, per-voxel semantic representations that are consistent across volumes, and we further employ off-the-shelf segmentation models to obtain coarse anatomical identities.
- Together, these representations drive a GPU-accelerated iterative registration solver (FireANTs) in a hybrid multistage framework: the first stage estimates alignment from dense multi-channel features, after which a second refines the deformation using both anatomix- and segmentation-based losses.
- Lastly, because registration should not introduce non-biological tumor volume changes across time, we develop lesion-specific component-wise volume preservation schemes.
Sources (1)
- [1]Registration is a Representation Problem: Solving Whole-Body PET/CT RegistrationarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 13, 05:45 PM
We present an approach that instead establishes correspondence using complementary representations of anatomy and demonstrate its utility for longitudinal whole-body PET/CT registration in the Learn2Reg 2026 PSMAReg Challenge.
Raw image intensities are a poor surrogate for similarity across images in correspondence problems.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 13, 2026NVIDIA Run:ai Is Now Open Source as Part of NVIDIA DSX AI Factory Platform
- Oct 13, 2026Fish Audio Drama 3 Technical Report
- Oct 13, 2026Many Right Answers, One Training Example: Removing Responder Noise from GRPO for LLM Recommenders
- Oct 13, 2026DeltaTTG: Training Multi-Turn Agents with a Token-to-Go Progress Meter
- Oct 13, 2026DisaggKD: Efficient Knowledge Distillation through Teacher-Student Disaggregation and a Fused Loss
- Oct 13, 2026Joint Subspace Sparse Attention for Efficient Long-Context Decoding