Table 6 Quantitative evaluation of component contributions in the proposed framework via ablation study.
From: Segmentation of gastroesophageal reflux events using a semi-U-Net architecture with 1D/2D CNNs
Configurations | Acc (%) | Sen (%) | PPV (%) | F1 | |
|---|---|---|---|---|---|
Name | Description | ||||
C1 | Without GER definition criteria | 93.12 ± 3.82 | 88.60 ± 4.85 | 73.62 ± 9.33 | 0.79 ± 0.06 |
C2 | Without morphological edit | 93.48 ± 1.18 | 66.41 ± 6.38 | 87.08 ± 6.90 | 0.74 ± 0.03 |
C3 | Without cross sections with BI | 92.37 ± 3.88 | 89.99 ± 4.78 | 70.15 ± 8.50 | 0.78 ± 0.05 |
C4 | No post-process | 93.57 ± 0.92 | 69.82 ± 7.08 | 84.60 ± 5.91 | 0.76 ± 0.02 |
C5 | Proposed method | 96.72 ± 0.88 | 84.90 ± 6.16 | 92.05 ± 1.71 | 0.88 ± 0.03 |
