Table 3 Performance analysis of the developed and existing segmentation CNNs (trained from scratch).
From: COVID-19 infection analysis framework using novel boosted CNNs and radiological images
Model | Region | DSC | Acc | IoU | BF | Global-Acc | Mean-Acc | Mean-IoU | Weighted-IoU | Mean-BF |
|---|---|---|---|---|---|---|---|---|---|---|
Ablation study of the proposed segmentation CNNs | ||||||||||
Proposed SA-CB-BRSeg | Infected | 96.40 | 99.21 | 98.85 | 99.09 | 99.51 | 99.49 | 98.98 | 99.09 | 98.32 |
Background | 99.02 | 99.72 | 99.31 | 97.45 | ||||||
Proposed CB-BRSeg | Infected | 95.96 | 99.01 | 98.43 | 98.87 | 99.25 | 99.17 | 98.69 | 98.80 | 98.03 |
Background | 98.90 | 99.48 | 99.09 | 97.33 | ||||||
Proposed SA-BRSeg | Infected | 95.61 | 98.83 | 98.35 | 98.47 | 98.99 | 98.97 | 98.46 | 98.57 | 97.42 |
Background | 98.40 | 99.38 | 98.85 | 96.73 | ||||||
Comparative studies with the existing CNNs | ||||||||||
Deeplabv3 | Infected | 95.00 | 98.48 | 97.59 | 97.53 | 99.03 | 98.91 | 98.14 | 98.33 | 97.08 |
Background | 98.30 | 99.33 | 98.67 | 96.39 | ||||||
nnSAM41 | Infected | 94.85 | 98.33 | 97.45 | 97.39 | 98.93 | 98.81 | 98.04 | 98.23 | 96.88 |
Background | 98.27 | 99.30 | 98.64 | 96.36 | ||||||
U-SegNet | Infected | 94.65 | 98.25 | 97.01 | 97.02 | 98.82 | 98.73 | 97.52 | 97.68 | 96.52 |
Background | 98.01 | 99.16 | 98.10 | 95.22 | ||||||
nnUNet | Infected | 94.63 | 98.20 | 96.09 | 96.98 | 98.80 | 98.69 | 97.49 | 97.65 | 96.49 |
Background | 98.27 | 99.18 | 98.12 | 95.27 | ||||||
SegNet | Infected | 94.30 | 98.97 | 96.56 | 96.73 | 98.72 | 98.71 | 97.18 | 97.32 | 96.48 |
Background | 97.90 | 98.45 | 97.79 | 95.22 | ||||||
U-Net | Infected | 94.00 | 98.61 | 95.98 | 96.91 | 98.40 | 98.44 | 96.70 | 96.87 | 95.49 |
Background | 97.70 | 98.28 | 97.42 | 94.07 | ||||||
VGG-16 | Infected | 91.00 | 91.38 | 88.91 | 89.37 | 95.59 | 94.81 | 91.05 | 91.53 | 83.66 |
Background | 95.00 | 98.25 | 93.18 | 77.95 | ||||||
FCN-8 | Infected | 90.70 | 90.92 | 89.11 | 87.74 | 95.32 | 94.55 | 90.63 | 90.20 | 82.43 |
Background | 94.00 | 98.18 | 92.15 | 77.11 | ||||||
Comparative studies with reported segmentation techniques | ||||||||||
VB-Net48 | Infected | 91.00 | – | – | –– | – | – | – | – | – |
Weakly Sup 50 | Infected | 90.00 | – | –– | – | – | – | – | – | – |
MTL51 | Infected | 88.00 | – | – | – | – | – | – | – | – |
DCN28 | Infected | 83.50 | – | – | – | – | – | – | – |  |
U-Net-CA30 | Infected | 83.10 | – | – | – | – | – | – | – | – |
Inf-Net52 | Infected | 68.20 | – | – | – | – | – | – | – | – |
