Table 3 Total area errors, precisions and sensitivities per each layer. Since there are two classes in the portal areas and parenchymal layer, highest class-specific errors have been evaluated for both classes. Total area errors are the sum of false positive (FP) and false negative (FN), namely the total errors per training areas in each layer of the AI model
Layer for Liver Tissue | Layer for Portal areas & Parenchymaa | Layer for K7 positive hepatocytes | |
|---|---|---|---|
Total area error (%) | 0.56 | 2.6 | 0.24 |
Precision of the segmentation (%) | 99.4 | 98.4 | 92.9 |
Sensitivity of the segmentation (%) | 99.6 | 98.3 | 91.8 |
Highest class-specific error accepted per class % (FP%/FN%1) | 1.02% (0.63%/0.39%) | 2.26% (1.03%/1,23%)2, 6.82% (3.42%/3.40%)3 | 15.24% (7.00%/8.24%) |