Introduction
Proposed method
Extending the core models for handling multi-frame information
Patient no. | Video no. | No. of annotated frames | Image size (pixels) |
---|---|---|---|
1 | Video 1 | 21 | 356 \(\times \) 256 |
1 | Video 2 | 240 | 256 \(\times \) 266 |
2 | Video 3 | 462 | 296 \(\times \) 277 |
2 | Video 4 | 234 | 296 \(\times \) 277 |
3 | Video 5 | 51 | 296 \(\times \) 277 |
4 | Video 6 | 201 | 296 \(\times \) 277 |
5 | Video 7 | 366 | 256 \(\times \) 262 |
6 | Video 8 | 387 | 256 \(\times \) 262 |
6 | Video 9 | 234 | 256 \(\times \) 262 |
6 | Video 10 | 117 | 256 \(\times \) 262 |
6 | Video 11 | 360 | 256 \(\times \) 262 |
Total | – | 2673 | – |
Evaluation
Dataset
Training setting
Performance metrics
Ablation study and comparison with state of the art
Results
F(*) | DSC | Prec | Rec |
---|---|---|---|
(\(m_1\), \(m_2\)) | 0.78 | 0.65 | 0.71 |
(\(M_1\), \(M_2\)) | 0.71 | 0.55 | 0.57 |
(\(M_1\), \(m_1\)) | 0.72 | 0.56 | 0.66 |
(\(M_2\), \(m_2\)) | 0.68 | 0.51 | 0.63 |