Reproduce “Leveraging Segmentation Representations for Efficient Classification: A Self-Configuring Framework Based on nnU-Net” 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

Renjun Cai, Antonio Javier Gallego, Félix Fuentes-Hurtado

The framework extends nnU-Net segmentation to scan-level classification by reusing encoder features and predicted mask probabilities in a small classification head, rather than training a second encoder. In five-fold EGCT evaluation, adding a segmentation-guided region-of-interest feature changed Overall from 0.7771 to 0.7826, with overlapping 95% confidence intervals. The submission ranked first overall across four challenge tasks, but did not lead every metric on every task.

Summary by alphaXiv.

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