Reproduce “EDMAE: An Efficient Decoupled Masked Autoencoder for Standard View Identification in Pediatric Echocardiography” 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
Yiman Liu, Xiaoxiang Han, Tongtong Liang, Bin Dong, Jiajun Yuan, Menghan Hu, Qiaohong Liu, Jiangang Chen and 2 more
EDMAE learns from unlabeled pediatric echocardiograms by aligning representations of visible and masked regions while a decoder reconstructs missing pixels. On a private test set of 29 view categories, it reports 98.48% accuracy and 93.63% mean F1, versus ConvMAE’s 98.45% and 92.97%. Reported pretraining convergence takes 24 hours, but these results establish neither clinical benefit nor generalization beyond this setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.