Research map / Federated learning and privacy / Personalized and heterogeneous FL
Shift-Dependent Asymmetry: Orthogonal Inverse Low-Rank Adaptation for Federated Medical Segmentation
Xingyue Zhao, Wenke Huang, Linghao Zhuang, Haoran Wu, Anwen Jiang, Zhifeng Wang, Wenwen He, Ming Feng, Mang Ye, Bo XU
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add Inverse Asymmetric Tuning (local-A/shared-B in encoder, shared-A/local-B in decoder) (ablation) | Federated LoRA on SAM · Fundus OD/OC avg | DSC: 77.33 -> 83.47 | Improved |
| Add Subspace Orthogonality Regularizer on top of IAT (ablation) | FedIAT · Fundus OD/OC avg | DSC: 83.47 -> 84.52 | Improved |
| Add IAT on DRISHTI-GS1 client (ablation) | Federated LoRA on SAM · Fundus DRISHTI-GS1 | DSC: 57.74 -> 81.64 | Improved |
| Add SOR on top of IAT on REFUGE and G1020 (ablation) | FedIAT · Fundus REFUGE and G1020 | DSC: REFUGE 89.13 -> 89.00; G1020 77.71 -> 78.15 | Mixed |
| Reverse allocation rule (opposite of IAT) (ablation) | FedIAT · Fundus OD/OC avg | DSC: 83.47 -> 82.54 | Hurt |
| Use inverse allocation instead of uniform sharing (FedSA) (ablation) | FedIAT · Fundus OD/OC avg | DSC: 83.04 -> 83.47 | Improved |
| Set SOR weight lambda to 1e-4 instead of negligible 1e-6 (ablation) | FedIAT · Histology and Fundus | DSC: 81.40 (Histology), 84.52 (Fundus) at lambda 1e-4 | Improved |
| Set SOR weight lambda to overly large 1e-3 (ablation) | FedIAT · Histology and Fundus | DSC | Hurt |
| Apply LoRA to encoder and decoder instead of encoder only (ablation) | LoRA on SAM · Medical segmentation (preliminary) | DSC | Improved |
| Freeze LoRA A factor (FFA-LoRA) instead of aggregating | Federated LoRA · Histology TNBC client | DSC: 3.59 on TNBC | Hurt |
| Increase LoRA rank from 8 to 16 | FedIAT · Histology nuclei avg | DSC: 81.40 -> 81.49 | No clear effect |
| Share only one LoRA factor instead of both | FedIAT · Per-round communication | Communicated params: 0.78M -> 0.55M | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FedSA — FedSA: A Unified Representation Learning via Semantic Anchors for Prototype-based Federated Learning (AAAI 2025) · better on 4 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- S2S2: Semantic Stacking for Robust Semantic Segmentation in Medical Imaging (AAAI 2025 · similarity 0.85)
- Gradient Alignment Improves Test-Time Adaptation for Medical Image Segmentation (AAAI 2025 · similarity 0.85)
- DyCON: Dynamic Uncertainty-aware Consistency and Contrastive Learning for Semi-supervised Medical Image Segmentation (CVPR 2025 · similarity 0.85)
- Convolution Meets LoRA: Parameter Efficient Finetuning for Segment Anything Model (ICLR 2024 · similarity 0.85)
- GapMatch: Bridging Instance and Model Perturbations for Enhanced Semi-Supervised Medical Image Segmentation (AAAI 2025 · similarity 0.85)
- Why Invariance is Not Enough for Biomedical Domain Generalization and How to Fix It (NeurIPS 2026 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
