Skip to main content
. Author manuscript; available in PMC: 2015 Nov 24.
Published in final edited form as: Conf Proc IEEE Eng Med Biol Soc. 2014;2014:194–197. doi: 10.1109/EMBC.2014.6943562

Table 2.

Average stain segmentation accuracy using color deconvolution normalization (KM, EM, and VB) and quantile normalization (AP and CM) methods.

Testing OV GBM RCC1 RCC2
Training Method Mean Mean Mean Mean
OV KM 83 ± 7.2 91 ± 6.4 80 ± 12.6
EM 83 ± 7.1 91 ± 6.4 80 ± 12.6
VB 83 ± 7.2 90 ± 6.5 80 ± 12.6
AP 80 ± 11.1 78 ± 9.4 74 ± 9.2
CM 85 ± 7.0 87 ± 7.9 84 ± 5.9
GBM KM 94 ± 6.5 92 ± 7.6 84 ± 10.0
EM 94 ± 6.5 93 ± 7.5 84 ± 10.0
VB 94 ± 6.5 92 ± 7.5 85 ± 10.0
AP 83 ± 7.2 84 ± 6.2 80 ± 7.6
CM 87 ± 5.8 87 ± 5.5 83 ± 6.1
RCC1 KM 92 ± 4.2 87 ± 6.7 91 ± 6.8
EM 92 ± 4.2 87 ± 6.7 91 ± 6.8
VB 92 ± 4.2 85 ± 7.0 91 ± 6.8
AP 83 ± 7.8 85 ± 7.9 82 ± 7.5
CM 81 ± 8.9 82 ± 9.2 82 ± 7.3
RCC2 KM 91 ± 4.5 87 ± 6.5 96 ± 9.2
EM 91 ± 4.5 87 ± 6.5 96 ± 9.2
VB 91 ± 4.5 87 ± 6.7 95 ± 9.3
AP 85 ± 7.0 84 ± 8.4 88 ± 6.0
CM 87 ± 5.2 87 ± 7.5 84 ± 11.4

Performance of KM methods are highlighted in bold red where either (1) performance is significantly better than all other methods for the particular test case, or (2) performance is not significantly different from any other methods. (Student’s t-tests, p-value<0.05).