Table 13.
Ablation study of the proposed multi-view perspective fusion strategy (MVP) using repeated single-view inputs with distinct positional embeddings.
| Method | mIoU | Dice |
|---|---|---|
| MVGFormer(XY*3) | 82.4 | 90.3 |
| MVGFormer(YZ*3) | 83.1 | 90.5 |
| MVGFormer(XZ*3) | 82.6 | 89.9 |
| MVGFormer(P3DA) | 86.9 | 93.1 |