Reproduce “LoRe-MIM: Lost-or-Replaced Masked Image Modeling for Self-Supervised Representation Learning in Whole-Body PET/CT” 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
Amartya Bhattacharya, Bashirul Azam Biswas, Biratal Raj Wagle, Michael S. Leapman, Matthew E. Maeder, James B. Yu, Indrani Bhattacharya
LoRe-MIM pretrains a model on paired PET/CT scans without lesion labels, reconstructing blocks that are erased or replaced with corresponding content from the other modality. On an external test of 200 metastatic prostate cancer scans, encoder-and-decoder fine-tuning achieved a Dice score of 0.613, versus 0.547 for training SwinUNETR from scratch; this result does not establish performance across cancers or prospective settings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.