Reproduce “Bayesian PET Reconstruction with Learned Flow Matching Priors via Langevin Sampling” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Hengjia Ran, Huafeng Liu, Bo Zhao
The method reconstructs low-dose PET images by learning a flow-matching generator of images and sampling its latent space with count constraints. It uses Langevin updates to produce posterior samples, whose average gives an MMSE image and whose spread estimates uncertainty. In a simulated 10%-dose phantom test, it reports lower NRMSE than DPS, 7.35% versus 8.29%; evidence is limited to simulation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.