Table 2.

Conditional probability of class assignment, proportions of agreement, and 95% CIs for reported experiments

Experiment DescriptionComparisonProportions of Agreement (95% CIs)
I (n=11)IIa (n=2)IIb (n=12)III (n=23)IV (n=6)
 RP1 versus GT, scheme: T, data: patientsProportion of agreement1 (0.03 to 1)0.07 (0.004 to 0.34)0.26 (0.13 to 0.45)0.63 (0.44 to 0.78)0.13 (0.007 to 0.53)
P(RP1|GT)1 (0.29 to 1)0.5 (0.01 to 0.99)0.4 (0.19 to 0.64)0.71 (0.51 to 0.87)0.13 (0.003 to 0.53)
 RP2 versus GT, scheme: T, data: patientsProportion of agreement1 (0.31 to 1)0.33 (0.02 to 0.87)0.64 (0.44 to 0.81)0.80 (0.61 to 0.92)0.33 (0.09 to 0.69)
P(RP2|GT)1 (0.29 to 1)0.50 (0.01 to 0.99)0.90 (0.68 to 0.99)0.86 (0.67 to 0.96)0.38 (0.09 to 0.76)
 C versus GT, scheme: T, data: patientsProportion of agreement0.67 (0.58 to 0.76)0.03 (0.005 to 0.12)0.11 (0.07 to 0.17)0.47 (0.42 to 0.52)0.04 (0.01 to 0.13)
P(C|GT)0.67 (0.58 to 0.76)0.10 (0.01 to 0.32)0.18 (0.12 to 0.26)0.76 (0.66 to 0.81)0.05 (0.01 to 0.14)
I (n=18)IIa (n=6)IIb (n=27)III (n=58)IV (n=12)
 C versus GT, scheme: T, data: sectionsProportion of agreement0.90 (0.80 to 0.96)0.69 (0.48 to 0.85)0.72 (0.63 to 0.79)0.9 (0.85 to 0.93)0.70 (0.60 to 0.85)
P(C|GT)0.90 (0.81 to 0.96)0.75 (0.53 to 0.90)0.84 (0.76 to 0.91)0.97 (0.94 to 0.99)0.83 (0.70 to 0.93)
Normal (n=2)Mild (n=18)Moderate (n=13)Severe (n=0)
 C versus GT, scheme: F, data: patientsProportion of agreement0 (0 to 0.37)0.52 (0.42 to 0.62)0.43 (0.31 to 0.55)X
P(C|GT)0 (0 to 0.26)0.75 (0.66 to 0.83)0.60 (0.49 to 0.71)
Normal (n=6)Mild (n=46)Moderate (n=33)Severe (n=0)
 C versus GT, scheme: F, data: sectionsProportion of agreement1 (0.78 to 1)0.97 (0.93 to 0.99)0.96 (0.89 to 0.99)X
P(C|GT)1 (0.84 to 1)1 (0.97 to 1)0.96 (0.90 to 0.99)
  • T refers to classifications according to the Tervaert scheme; F refers to classifications according to the Fogo scheme. Data description identifies whether the experiment was performed using separate patients or separate sections as individual data. Each value is reported with the format observed value (lower 95% CI to upper 95% CI). Columns also indicate the number of samples in each class. GT, ground truth; RP, renal pathologist; P(RP1|GT), conditional probability of the class selection by the RP1 given the GT class; P(C|GT), conditional probability of the class prediction by the C given the GT class; X, not applicable.