Research Article
Robust Blood Cell Image Segmentation Method Based on Neural Ordinary Differential Equations
Table 1
Comparison results of different ODE-block positions.
| Algorithm | PA (%) | CPA (%) | MIoU (%) | Time (s) | Background | Red cells | White cells |
| U-Net | 94.59 | 92.55 | 96.54 | 94.10 | 88.58 | 0.15 | ODE-UNet1 | 95.05 | 94.25 | 95.97 | 93.92 | 89.80 | 7.35 | ODE-UNet2 | 95.08 | 94.23 | 95.69 | 95.98 | 89.68 | 3.73 | ODE-UNet3 | 95.11 | 95.14 | 95.70 | 91.65 | 89.61 | 1.94 | ODE-UNet4 | 95.19 | 94.08 | 96.51 | 93.42 | 90.17 | 1.03 | ODE-UNet5 | 95.05 | 95.62 | 94.62 | 94.53 | 89.62 | 0.61 | ODE-UNet6 | 95.14 | 93.78 | 96.34 | 95.39 | 90.09 | 1.06 | ODE-UNet7 | 95.17 | 93.80 | 96.59 | 94.19 | 90.02 | 1.95 | ODE-UNet8 | 95.16 | 94.33 | 96.30 | 92.98 | 89.91 | 3.72 | ODE-UNet9 | 95.14 | 94.33 | 96.23 | 93.15 | 90.03 | 7.35 |
|
|