Survival Prediction on TCGA-BRCA
0.76C-indexGrapHist
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GrapHist2026.02 | 0.76 | 1.29 | — | |
| PANTHERAll + MLPLearning Paradigm=Ours, Features=UNI, Prototypes (C)=16, Evaluation Protocol=MLP2024.05 | 0.758 | — | — | |
| OTLearning Paradigm=Unsupervised, Features=UNI, Prototypes (C)=162024.05 | 0.755 | — | — | |
| MMPInference-time modality=Multimodal inference: WSI+omics at test time2026.06 | 0.7423 | — | — | |
| STEPHTraining Source=Model merging-based knowledge transfer from Ms2026.03 | 0.7408 | — | — | |
| PANTHERAll + lin.Learning Paradigm=Ours, Features=UNI, Prototypes (C)=16, Evaluation Protocol=Linear2024.05 | 0.722 | — | — | |
| MAE2026.02 | 0.72 | 1.54 | — | |
| ROUPKTTraining Source=Representation-based knowledge transfer from Ms2026.03 | 0.7181 | — | — | |
| MurreNetPathological images=true, Genomic profiles=true2025.07 | 0.718 | — | — | |
| PANTHERTop + lin.Learning Paradigm=Ours, Features=UNI, Prototypes (C)=16, Evaluation Protocol=Linear2024.05 | 0.718 | — | — | |
| HFGPIG. (Genomic)=✓, P. (Proteomic)=✓, I. (Pathology Imaging)=✓2026.03 | 0.715 | — | — | |
| ReconMILFeature Extractor=ResNet-502026.03 | 0.71 | — | — | |
| ReconMILFeature Extractor=PLIP2026.03 | 0.709 | — | — | |
| MambaFeature Extractor=ResNet-502026.03 | 0.704 | — | — | |
| PS3†G. (Genomic)=✓, P. (Proteomic)=✓, I. (Pathology Imaging)=✓2026.03 | 0.702 | — | — | |
| LD-CVAEInference-time modality=Multimodal inference: WSI+omics at test time2026.06 | 0.6994 | — | — | |
| WiKGI. (Pathology Imaging)=✓2026.03 | 0.699 | — | — | |
| DMMLtInference-time modality=Multimodal inference: WSI+omics at test time2026.06 | 0.6986 | — | — | |
| ABMILI. (Pathology Imaging)=✓2026.03 | 0.698 | — | — | |
| SurvPathG. (Genomic)=✓, I. (Pathology Imaging)=✓2026.03 | 0.694 | — | — | |
| PAMTInference-time modality=Multimodal inference: WSI+omics at test time2026.06 | 0.6935 | — | — | |
| ICFNet†G. (Genomic)=✓, P. (Proteomic)=✓, I. (Pathology Imaging)=✓2026.03 | 0.692 | — | — | |
| MoPEInference-time modality=Knowledge distillation: WSI+omics training, WSI-only inference2026.06 | 0.6918 | — | — | |
| MCATPathological images=true, Genomic profiles=true2025.07 | 0.691 | — | — | |
| TransMILI. (Pathology Imaging)=✓2026.03 | 0.691 | — | — | |
| DeltaMILFeature Extractor=UNI [7]2025.12 | 0.6898 | — | — | |
| DeltaMILFeature Extractor=UNI2025.12 | 0.6898 | — | — | |
| PIBDG. (Genomic)=✓, I. (Pathology Imaging)=✓2026.03 | 0.689 | — | — | |
| MoMEG. (Genomic)=✓, I. (Pathology Imaging)=✓2026.03 | 0.688 | — | — | |
| MOTCatG. (Genomic)=✓, I. (Pathology Imaging)=✓2026.03 | 0.686 | — | — | |
| TopoMamSurv2026.05 | 0.686 | — | — | |
| SurvPathPathological images=true, Genomic profiles=true2025.07 | 0.685 | — | — | |
| MCATG. (Genomic)=✓, I. (Pathology Imaging)=✓2026.03 | 0.685 | — | — | |
| ReconMILFeature Extractor=CONCH v1.52026.03 | 0.682 | — | — | |
| CMTAG. (Genomic)=✓, I. (Pathology Imaging)=✓2026.03 | 0.681 | — | — | |
| RankByGeneImage Encoder=RankByGene2024.11 | 0.681 | — | — | |
| TIES AMTraining Source=Model merging-based knowledge transfer from Ms2026.03 | 0.6799 | — | — | |
| CLAM-SBI. (Pathology Imaging)=✓2026.03 | 0.679 | — | — | |
| Mean-PoolingFeature Extractor=UNI [7]2025.12 | 0.6776 | — | — | |
| Mean-PoolingFeature Extractor=UNI2025.12 | 0.6776 | — | — | |
| MOTCatPathological images=true, Genomic profiles=true2025.07 | 0.675 | — | — | |
| CLAM-MBI. (Pathology Imaging)=✓2026.03 | 0.675 | — | — | |
| DeepSetsLearning Paradigm=Unsupervised, Features=UNI2024.05 | 0.673 | — | — | |
| Fine-tunedTraining Source=Traditional cancer-specific training on Mt2026.03 | 0.6723 | — | — | |
| H2TLearning Paradigm=Unsupervised, Features=UNI, Prototypes (C)=162024.05 | 0.672 | — | — | |
| BLEEPImage Encoder=BLEEP2024.11 | 0.672 | — | — | |
| MMPTrans.G. (Genomic)=✓, I. (Pathology Imaging)=✓2026.03 | 0.671 | — | — | |
| LWRDownstream predictor=penalized Cox's proportional hazards models, Evaluation protocol=five-fold cross-validation, Input features=patient information and embeddings2026.06 | 0.671 | — | 1 | |
| S4MILFeature Extractor=UNI [7]2025.12 | 0.6709 | — | — | |
| S4MILFeature Extractor=UNI2025.12 | 0.6709 | — | — | |
| PANTHERWA + lin.Learning Paradigm=Ours, Features=UNI, Prototypes (C)=16, Evaluation Protocol=Linear2024.05 | 0.67 | — | — | |
| UNIImage Encoder=UNI2024.11 | 0.668 | — | — | |
| CMTAPathological images=true, Genomic profiles=true2025.07 | 0.667 | — | — | |
| Protein exp.P. (Proteomic)=✓2026.03 | 0.666 | — | — | |
| VanillaTraining Source=Traditional cancer-specific training on Mt2026.03 | 0.6648 | — | — | |
| GraphMamba2026.05 | 0.664 | — | — | |
| GPNetInference-time modality=Gene-only inference2026.06 | 0.6619 | — | — | |
| G-HANetInference-time modality=Knowledge distillation: WSI+omics training, WSI-only inference2026.06 | 0.6604 | — | — | |
| MKDInference-time modality=Knowledge distillation: WSI+omics training, WSI-only inference2026.06 | 0.6601 | — | — | |
| DeltaMILFeature Extractor=ResNet-502025.12 | 0.6593 | — | — | |
| MOMEPathological images=true, Genomic profiles=true2025.07 | 0.659 | — | — | |
| TDCInference-time modality=Knowledge distillation: WSI+omics training, WSI-only inference2026.06 | 0.6582 | — | — | |
| DSMILPathological images=true, Genomic profiles=false2025.07 | 0.657 | — | — | |
| HEST-FTImage Encoder=HEST-FT2024.11 | 0.653 | — | — | |
| MINDDownstream predictor=penalized Cox's proportional hazards models, Evaluation protocol=five-fold cross-validation, Input features=patient information and embeddings2026.06 | 0.65 | — | 2 | |
| ILRALearning Paradigm=Supervised, Features=UNI2024.05 | 0.649 | — | — | |
| ABMILFeature Extractor=UNI [7]2025.12 | 0.649 | — | — | |
| ABMILFeature Extractor=UNI2025.12 | 0.649 | — | — | |
| IntegrAODownstream predictor=penalized Cox's proportional hazards models, Evaluation protocol=five-fold cross-validation, Input features=patient information and embeddings2026.06 | 0.646 | — | 3 | |
| DMMLInference-time modality=Knowledge distillation: WSI+omics training, WSI-only inference2026.06 | 0.6458 | — | — | |
| TransMIL-MoEInference-time modality=Histology-only inference2026.06 | 0.6443 | — | — | |
| ABMILLearning Paradigm=Supervised, Features=UNI2024.05 | 0.644 | — | — | |
| ABMILPathological images=true, Genomic profiles=false2025.07 | 0.642 | — | — | |
| CTransPathImage Encoder=CTransPath2024.11 | 0.641 | — | — | |
| TransMIL2026.05 | 0.641 | — | — | |
| CLAMFeature Extractor=UNI [7]2025.12 | 0.6401 | — | — | |
| CLAMFeature Extractor=UNI2025.12 | 0.6401 | — | — | |
| JASMINEDownstream predictor=penalized Cox's proportional hazards models, Evaluation protocol=five-fold cross-validation, Input features=patient information and embeddings2026.06 | 0.64 | — | 4 | |
| ASMILFeatures=CONCH2026.03 | 0.6396 | — | — | |
| WIKGInference-time modality=Histology-only inference2026.06 | 0.6395 | — | — | |
| M3IFPathological images=true, Genomic profiles=true2025.07 | 0.639 | — | — | |
| HGT2026.05 | 0.639 | — | — | |
| MambaMILFeature Extractor=UNI [7]2025.12 | 0.6387 | — | — | |
| MambaMILFeature Extractor=UNI2025.12 | 0.6387 | — | — | |
| Patch-GCNFeatures=CONCH2026.03 | 0.6375 | — | — | |
| MambaMILFeature Extractor=ResNet-502025.12 | 0.6364 | — | — | |
| Gene exp.G. (Genomic)=✓2026.03 | 0.635 | — | — | |
| TransMILFeature Extractor=UNI [7]2025.12 | 0.6345 | — | — | |
| TransMILFeature Extractor=UNI2025.12 | 0.6345 | — | — | |
| MSNEDownstream predictor=penalized Cox's proportional hazards models, Evaluation protocol=five-fold cross-validation, Input features=patient information and embeddings2026.06 | 0.634 | — | 5 | |
| DINOv22026.02 | 0.63 | 9.8 | — | |
| AttMIL-MoEInference-time modality=Histology-only inference2026.06 | 0.6274 | — | — | |
| AttnMISLLearning Paradigm=Supervised, Features=UNI, Prototypes (C)=162024.05 | 0.627 | — | — | |
| Mamba2MILFeature Extractor=ResNet-502025.12 | 0.6254 | — | — | |
| DTFD-MILPathological images=true, Genomic profiles=false2025.07 | 0.625 | — | — | |
| S4MILFeature Extractor=ResNet-502025.12 | 0.6248 | — | — | |
| PORPOISEPathological images=true, Genomic profiles=true2025.07 | 0.624 | — | — | |
| MambaBack2026.04 | 0.623 | — | — | |
| DSMILFeature Extractor=ResNet-502025.12 | 0.6223 | — | — | |
| SNNPathological images=false, Genomic profiles=true2025.07 | 0.621 | — | — |