Table 4 Statistical analysis of nuclear classification performance with 95% confidence intervals on TCGA-based Cell Classification Dataset
From: Instance-level semantic segmentation of nuclei based on multimodal structure encoding
Method | \(F_{\text {c}\_\text {avg}}\) | \(F_{\text {immune}}\) | \(F_{\text {stroma}}\) | \(F_{\text {tumor}}\) |
|---|---|---|---|---|
HoVer-Net [53] | 0.821 (0.814, 0.828) | 0.785 (0.778, 0.792) | 0.838 (0.831, 0.845) | 0.840 (0.833, 0.847) |
Mask2former [54] | 0.835 (0.828, 0.842) | 0.798 (0.791, 0.805) | 0.852 (0.845, 0.859) | 0.855 (0.848, 0.862) |
UNet++ [57] | 0.808 (0.801, 0.815) | 0.772 (0.765, 0.779) | 0.825 (0.818, 0.832) | 0.827 (0.820, 0.834) |
Triple U-net [55] | 0.808 (0.801, 0.815) | 0.772 (0.765, 0.779) | 0.825 (0.818, 0.832) | 0.827 (0.820, 0.834) |
DeepLabV3 [58] | 0.810 (0.803, 0.817) | 0.780 (0.773, 0.787) | 0.830 (0.823, 0.837) | 0.835 (0.828, 0.842) |
TSFD-net [56] | 0.815 (0.808, 0.822) | 0.779 (0.772, 0.786) | 0.832 (0.825, 0.839) | 0.834 (0.827, 0.841) |
PSPNet [59] | 0.806 (0.799, 0.813) | 0.775 (0.768, 0.782) | 0.818 (0.811, 0.825) | 0.820 (0.813, 0.827) |
SegNet [60] | 0.799 (0.792, 0.806) | 0.760 (0.753, 0.767) | 0.811 (0.804, 0.818) | 0.815 (0.808, 0.822) |
Ours | 0.858 (0.851, 0.865) | 0.822 (0.815, 0.829) | 0.875 (0.868, 0.882) | 0.877 (0.870, 0.884) |