Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
SGCD: Stain-Guided CycleDiffusion for Unsupervised Domain Adaptation of Histopathology Image Classification
Authors: Hsi-Ling Chen, Chun-Shien Lu, Pau-Choo Chung
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments conducted on four public datasets demonstrate that SGCD can effectively enhance the performance of downstream task models on the target domain. ... Experiments were conducted on Camelyon17-WILDS. Table 2 presents the comparison results. ... Ablation Studies |
| Researcher Affiliation | Academia | Hsi-Ling Chen Miin-Wu School of Computing National Cheng Kung University, Taiwan EMAIL Chun-Shien Lu Institute of Information Science Academia Sinica, Taiwan EMAIL Pau-Choo Chung Department of Electrical Engineering National Cheng Kung University & National Chung Cheng University, Taiwan EMAIL |
| Pseudocode | Yes | A.1 Pseudo Code of Proposed Method-SGCD Algorithms 1 and 2 show the pseudo-codes of the proposed SGCD method. Algorithm 1 Reverse Process Guided by Conditions Algorithm 2 S T S conversion of SGCD |
| Open Source Code | No | We do not include the code in this submission, but in the future, we will organize the experimental code and documentation, and released on Git Hub. |
| Open Datasets | Yes | SGCD was evaluated on four open datasets: Camelyon17 Bejnordi et al. (2017), Camelyon16 Bejnordi et al. (2017), Camelyon17-WILDS Koh et al. (2021), and MITOS & ATYPIA14 Racoceanu et al. (2014). The details of the four datasets are provided in Sec. A.2 of the Appendix. |
| Dataset Splits | Yes | MITOS & ATYPIA14 is obtained from the same slide samples scanned by two scanners, namely Aperio Scanscope XT (A) and Hamamatsu Nanozoomer 2.0-HT (H). A training set was constructed consisting of 10, 000 patches randomly selected from the first 184 WSIs of the two scanners. Furthermore, 500 patches from the remaining 100 WSIs from the scanners were selected at random for testing. |
| Hardware Specification | Yes | The experiments were implemented on NVIDIA V100 GPU with Python 3.10.12 and Pytorch 2.4.0. |
| Software Dependencies | Yes | The experiments were implemented on NVIDIA V100 GPU with Python 3.10.12 and Pytorch 2.4.0. |
| Experiment Setup | Yes | The Adam optimizer was employed with a learning rate of 2e 4 and batch size of 4. The total timestep T of the diffusion models was set to 1000. Stain guidance was applied from timestep 600 to 100. |