Medical Image Segmentation on Federated Medical Datasets (Average)
82.9Dice ScoreDSFedMed
Evaluation Results
| Method | Links | |
|---|---|---|
| DSFedMedInference Model=TinySAM, Communication Overhead=8,920 MB, Inference Time (per image)=0.015 s, Synchronous Training Time=198.02 min, Asynchronous Training Time=455.5 min2026.01 | 82.9 | |
| FedMSAInference Model=MSA-B, Communication Overhead=11,230 MB, Inference Time (per image)=0.127 s, Synchronous Training Time=1,160.00 min2026.01 | 81.1 | |
| FedSAMInference Model=SAM-B, Communication Overhead=71,538 MB, Inference Time (per image)=0.118 s, Synchronous Training Time=1,411.67 min2026.01 | 81 | |
| FedTinySAMInference Model=TinySAM, Communication Overhead=7,770 MB, Inference Time (per image)=0.015 s, Synchronous Training Time=198.02 min2026.01 | 79.1 | |
| FednnU-NetInference Model=nnU-Net, Communication Overhead=2,690 MB, Inference Time (per image)=0.018 s, Synchronous Training Time=345.68 min2026.01 | 75.9 | |
| FedU-NetInference Model=U-Net, Communication Overhead=2,430 MB, Inference Time (per image)=0.007 s, Synchronous Training Time=59.50 min2026.01 | 73.2 |