Skip to main content
. 2024 Feb 26;37(4):1691–1710. doi: 10.1007/s10278-024-01041-w

Table 6.

Evaluation of deep features combined with traditional algorithms from the literature and stain normalization on non-segmented images

LOG RF SVM
Method Reference ACC(%) F1 ACC(%) F1 ACC(%) F1
No normalization - 86.44 0.83 90.93 0.88 87.73 0.82
Macenko et al. [42] Ref-1 86.62 0.83 91.17 0.88 89.84 0.86
Ref-2 88.86 0.85 92.26 0.89 89.33 0.86
Ref-3 88.94 0.85 91.23 0.88 90.42 0.88
Reinhard et al. [43] Ref-1 80.33 0.78 83.22 0.81 82.12 0.79
Ref-2 84.13 0.81 85.15 0.83 89.78 0.84
Ref-3 83.28 0.81 85.83 0.84 85.34 0.81
Tosta et al. [28] Ref-1 85.70 0.83 87.96 0.85 87.96 0.85
Ref-2 88.53 0.85 89.34 0.86 88.73 0.85
Ref-3 83.31 0.81 85.64 0.83 85.91 0.82
Vahadane et al. [27] Ref-1 85.63 0.82 87.04 0.84 86.29 0.83
Ref-2 86.07 0.84 86.92 0.84 88.12 0.85
Ref-3 87.13 0.84 89.08 0.86 89.23 0.86