Lesion Detection on NIH DeepLesion (test)
82.22Sensitivity @ 0.5 FPsSVD-Net
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SVD-NetBackbone=P3D-63, Slices=9 slices, Pre-training=MS-COCO2022.01 | 82.22 | 87.42 | 90.91 | 93.65 | 88.55 | |
| MP3DBackbone=MP3D-63, Slices=9 slices2022.01 | 79.6 | 85.29 | 89.61 | 92.45 | 86.74 | |
| AlignShiftBackbone=DenseNet3D-121, Slices=7 slices2022.01 | 79.4 | 85.5 | 90.09 | 93.26 | 87.06 | |
| SVD-NetBackbone=ResNet3DV1c-18, Slices=9 slices, Pre-training=MS-COCO2022.01 | 79.28 | 84.8 | 89.04 | 91.9 | 86.26 | |
| ACSBackbone=DenseNet3D-121, Slices=7 slices2022.01 | 78.38 | 85.39 | 90.07 | 93.19 | 86.76 | |
| SVD-NetBackbone=ResNet3D-18, Slices=9 slices, Pre-training=MS-COCO2022.01 | 76.07 | 82.16 | 86.67 | 90.12 | 83.76 | |
| SVD-NetBackbone=ResNet3D-18, Slices=9 slices, Pre-training=ImageNet2022.01 | 74.18 | 81.55 | 86.77 | 90.68 | 83.3 | |
| MVP-NetBackbone=ResNet-50, Slices=9 slices2022.01 | 73.83 | 81.82 | 87.6 | 91.3 | 83.64 | |
| KineticsBackbone=ResNet3D-18, Slices=9 slices, Pre-training=Kinetics2022.01 | 72.94 | 80.92 | 86 | 89.91 | 82.44 | |
| ACSBackbone=ResNet3D-18, Slices=9 slices2022.01 | 71.16 | 78.95 | 84.98 | 89.2 | 81.07 | |
| ScratchBackbone=ResNet3D-18, Slices=9 slices, Pre-training=None2022.01 | 66.51 | 74.2 | 80.33 | 85.22 | 76.57 | |
| Med3DBackbone=ResNet3D-18, Slices=9 slices, Pre-training=Med3D2022.01 | 58.66 | 68.36 | 76.11 | 82.46 | 71.4 | |
| I3DBackbone=ResNet3D-18, Slices=9 slices, Pre-training=I3D2022.01 | 57.93 | 67.51 | 75.54 | 81.69 | 70.67 | |
| 3DCEBackbone=ResNet-50, Slices=27 slices2022.01 | 52.86 | 64.8 | 74.84 | 84.38 | 69.22 |