Reproduce “ADNP-15: An Open-Source Histopathological Dataset for Neuritic Plaque Segmentation in Human Brain Whole Slide Images with Frequency Domain Image Enhancement for Stain Normalization” 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
Chenxi Zhao, Jianqiang Li, Qing Zhao, Jing Bai, Susana Boluda, Benoit Delatour, Lev Stimmer, Daniel Racoceanu and 2 more
Researchers introduced ADNP-15, an open-source dataset containing 4000 expert-annotated neuritic plaques from 15 Alzheimer's patients in human brain whole slide images. They proposed a frequency-domain image enhancement method that consistently improved deep learning segmentation performance, with models like Dual-UNet achieving a 74.02% Dice score and DynUNet reaching a 76.74% F1-score when combined with specific stain normalization.
Key insights
- The frequency-domain image enhancement consistently improved segmentation performance across most deep learning models and stain normalization methods, increasing Dice/F1 scores and reducing MASD.
- DynUNet and Dual-UNet demonstrated the most substantial performance gains after applying the proposed enhancement method, reinforcing their robustness.
- Vahadane and Reinhard normalization methods generally yielded higher segmentation performance compared to Macenko and HistomicsTK, especially when combined with the proposed enhancement.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.