Lung nodule malignancy prediction on NLST (test)
93.1AUC (All)PARE
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| PARELearning paradigm=multi-task learning, Ensemble of deep supervision heads=true2023.07 | 93.1 | 89 | 78.1 | 82.7 | — | |
| PARELearning paradigm=multi-task learning2023.07 | 92.8 | 88.2 | 77 | 82.6 | — | |
| CA-NetLearning paradigm=multi-task learning2023.07 | 91.6 | 83.3 | 75.9 | 80.7 | — | |
| nnUnetLearning paradigm=pure segmentation2023.07 | 91 | 81.5 | 73.6 | 81.5 | — | |
| MiTLearning paradigm=pure classification2023.07 | 90.8 | 82.1 | 75.5 | 81 | — | |
| ASPPLearning paradigm=pure classification2023.07 | 90.2 | 79.8 | 71.6 | 80.1 | — | |
| CNNLearning paradigm=pure classification2023.07 | 89.4 | 74.2 | 70.6 | 78 | — | |
| Real follow-up LDCTMethod category=Real Image Baselines2026.03 | 81.9 | — | — | — | 39.3 | |
| NAMD (Ours)Method category=Stochastic Methods, Number of stochastic samples=202026.03 | 80.5 | — | — | — | 34.6 | |
| ImageFlowNet (SDE)Method category=Stochastic Methods, Sampling algorithm=SDE, Number of stochastic samples=202026.03 | 78.2 | — | — | — | 33.9 | |
| CorrFlowNet (ODE)Method category=Deterministic Methods, Sampling algorithm=ODE2026.03 | 77.9 | — | — | — | 31.8 | |
| CorrFlowNet (SDE)Method category=Stochastic Methods, Sampling algorithm=SDE, Number of stochastic samples=202026.03 | 77.6 | — | — | — | 32.1 | |
| NAMD (w/o Lalign)Method category=Stochastic Methods, Ablation=without alignment loss L_align, Number of stochastic samples=202026.03 | 76.5 | — | — | — | 30.5 | |
| McWGANMethod category=Deterministic Methods2026.03 | 76.3 | — | — | — | 30.7 | |
| AdaFuseInput Modalities=CT, Clinical, Text, Modality Count=3, Fusion Strategy=Adaptive (AdaFuse), MFLOPs=1.1642026.01 | 76.2 | — | — | — | — | |
| ABC-tensorInput Modalities=CT, Clinical, Text, Modality Count=3, Fusion Strategy=tensor fusion, MFLOPs=1.7902026.01 | 75.9 | — | — | — | — | |
| AB-concatInput Modalities=CT, Clinical, Modality Count=2, Fusion Strategy=concatenation, MFLOPs=0.5592026.01 | 75.8 | — | — | — | — | |
| AB-meanInput Modalities=CT, Clinical, Modality Count=2, Fusion Strategy=mean pooling, MFLOPs=0.5572026.01 | 75.5 | — | — | — | — | |
| DynMMInput Modalities=CT, Clinical, Text, Modality Count=3, Fusion Strategy=Adaptive (DynMM), MFLOPs=1.6352026.01 | 75.4 | — | — | — | — | |
| ABC-meanInput Modalities=CT, Clinical, Text, Modality Count=3, Fusion Strategy=mean pooling, MFLOPs=1.6222026.01 | 74.8 | — | — | — | — | |
| AC-meanInput Modalities=CT, Text, Modality Count=2, Fusion Strategy=mean pooling, MFLOPs=1.6082026.01 | 74.5 | — | — | — | — | |
| MoEInput Modalities=CT, Clinical, Text, Modality Count=3, Fusion Strategy=Adaptive (MoE), MFLOPs=3.4922026.01 | 74.2 | — | — | — | — | |
| Real baseline LDCTMethod category=Real Image Baselines2026.03 | 74.2 | — | — | — | 26.3 | |
| AC-tensorInput Modalities=CT, Text, Modality Count=2, Fusion Strategy=tensor fusion, MFLOPs=1.4772026.01 | 73.9 | — | — | — | — | |
| AB-tensorInput Modalities=CT, Clinical, Modality Count=2, Fusion Strategy=tensor fusion, MFLOPs=0.4332026.01 | 73.5 | — | — | — | — | |
| ABC-concatInput Modalities=CT, Clinical, Text, Modality Count=3, Fusion Strategy=concatenation, MFLOPs=1.6262026.01 | 73.5 | — | — | — | — | |
| AC-concatInput Modalities=CT, Text, Modality Count=2, Fusion Strategy=concatenation, MFLOPs=1.6102026.01 | 73.3 | — | — | — | — | |
| A (CT)Input Modalities=CT image, Modality Count=1, MFLOPs=0.5432026.01 | 73.2 | — | — | — | — | |
| SD1.5Method category=Stochastic Methods, Number of stochastic samples=202026.03 | 70.1 | — | — | — | 24.5 | |
| DDL-CXRMethod category=Stochastic Methods, Number of stochastic samples=202026.03 | 70.1 | — | — | — | 23.7 | |
| BC-tensorInput Modalities=Clinical, Text, Modality Count=2, Fusion Strategy=tensor fusion, MFLOPs=1.0882026.01 | 68.5 | — | — | — | — | |
| BC-meanInput Modalities=Clinical, Text, Modality Count=2, Fusion Strategy=mean pooling, MFLOPs=1.0822026.01 | 67.8 | — | — | — | — | |
| Flux (Rectified Flow)Method category=Stochastic Methods, Sampling algorithm=Rectified Flow, Number of stochastic samples=202026.03 | 66.8 | — | — | — | 22.7 | |
| B (Clinical)Input Modalities=Clinical variables, Modality Count=1, MFLOPs=0.0172026.01 | 66.2 | — | — | — | — | |
| BC-concatInput Modalities=Clinical, Text, Modality Count=2, Fusion Strategy=concatenation, MFLOPs=1.0842026.01 | 66.1 | — | — | — | — | |
| C (Text)Input Modalities=Text reports, Modality Count=1, MFLOPs=1.0672026.01 | 57.6 | — | — | — | — |