Reproduce “MedVTok: A General-Purpose Medical Visual Tokenizer” 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

Chenglong Ma, Yuanfeng Ji, Junzhi Ning, Jiyao Liu, Yujie Wei, Shaohao Rui, Chaoyang Zhang, Wenjie Li and 13 more

MedVTok uses one tokenizer for medical images and volumes, combining slice-continuity training with alignment to multiple modality-specific experts. It is evaluated across six task types and 24 benchmarks. With the tokenizer frozen and a linear head trained, it averages 76.07% Dice across eight segmentation benchmarks, versus 71.99% for MedITok; this measures feature transfer, not standalone clinical performance.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.