Federated Pseudo-Labeling: A Data-Centric, Privacy-Preserving Framework for Medical Image Segmentation
Bibliographic record
Abstract
Although essential in the medical domain, protecting patient privacy often restricts data sharing across institutions. Moreover, publicly available datasets often suffer from poor and inconsistent annotation-particularly for image segmentation that requires precise pixel- or voxel-level annotations. As a result, deep learning models are frequently trained on single-institution datasets that are small and lack the heterogeneity of broader patient populations, which limits their generalizability. Federated learning (FL) enables collaborative model training across institutions by sharing model parameters instead of raw medical data. However, it requires uniform model architectures, which may not align with local hardware or software, and still exposes privacy risks through parameter sharing. Further, coordination across institutions with varying data volumes and annotation standards remains challenging, and exchanging model weights-especially for large models-is costly and slow. To address these limitations, we propose DCFed, a data-centric, semi-supervised framework that avoids sharing private data and model parameters by leveraging pseudo-labeling and uncertainty estimation on publicly available unannotated datasets. In our experiments, we use a modified U-Net with residual blocks, atrous spatial pyramid pooling, and convolutional block attention modules at the client level. DCFed improves performance by up to 8.9% on a breast cancer ultrasound dataset and 3.7% on a skin cancer dermoscopy dataset over local training. Notably, DCFed outperforms conventional FL methods such as FedAvg and FedNova across multiple clients in both tasks. In conclusion, DCFed surpasses both centralized training on local datasets and parameter-sharing FL approaches across institutions, establishing a scalable and privacy-preserving solution for real-world medical image segmentation.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.005 | 0.031 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.002 |
| Open science | 0.014 | 0.018 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; both teacher heads agree on what is shown here.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".