Table 5. Confusion matrix for eight-class segmentation using U-Net.

	Interstitium	Glomeruli	Proximal tubules	Distal tubules	Arteries	Tubulitis	Degenerated tubules	Atrophic tubules
Interstitium	0.82	0.015	0.077	0.036	0.0024	0.011	0.029	0.014
Glomeruli	0.083	0.85	0.030	0.013	0.0023	0.0050	0.012	0.0053
Proximal tubules	0.13	0.0030	0.70	0.033	0.00067	0.017	0.11	0.011
Distal tubules	0.14	0.0050	0.067	0.67	0.00096	0.077	0.017	0.015
Arteries	0.60	0.086	0.020	0.036	0.14	0.023	0.039	0.064
Tubulitis	0.21	0.00071	0.087	0.15	0.027	0.28	0.15	0.12
Degenerated tubules	0.16	0.0055	0.17	0.015	0.0033	0.065	0.52	0.054
Atrophic tubules	0.17	0.0026	0.063	0.035	0.0012	0.085	0.11	0.53

The ground truth labels are given vertically, and the segmentation model’s predictions are given horizontally.