Table 3 The areas under the receiver operating characteristic curve of radiomics models to predict proliferative hepatocellular carcinoma
Model | Training cohort | Internal test cohort | External test cohort |
|---|---|---|---|
Plain radiomics model | |||
Decision tree | 0.68 (0.62, 0.74) | 0.55 (0.45, 0.64) | 0.56 (0.45, 0.66) |
Logistic regression | 0.67 (0.59, 0.76) | 0.50 (0.36, 0.63) | 0.62 (0.49, 0.75) |
Random forest | 1.00 (1.00, 1.00) | 0.60 (0.47, 0.72) | 0.63 (0.51, 0.75) |
Arterial radiomics model | |||
Decision tree | 0.76 (0.70, 0.82) | 0.50 (0.38, 0.63) | 0.52 (0.39, 0.64) |
Logistic regression | 0.70 (0.62, 0.77) | 0.71 (0.60, 0.83) | 0.68 (0.56, 0.81) |
Random forest | 1.00 (1.00, 1.00) | 0.59 (0.47, 0.71) | 0.59 (0.46, 0.72) |
Venous radiomics model | |||
Decision tree | 0.77 (0.70, 0.83) | 0.57 (0.44, 0.69) | 0.65 (0.54, 0.76) |
Logistic regression | 0.66 (0.58, 0.74) | 0.63 (0.49, 0.76) | 0.59 (0.47, 0.72) |
Random forest | 1.00 (1.00, 1.00) | 0.55 (0.42, 0.68) | 0.54 (0.40, 0.68) |
Delayed radiomics model | |||
Decision tree | 0.73 (0.65, 0.79) | 0.44 (0.33, 0.57) | 0.52 (0.40, 0.63) |
Logistic regression | 0.69 (0.60, 0.78) | 0.62 (0.47, 0.75) | 0.56 (0.44, 0.69) |
Random forest | 1.00 (1.00, 1.00) | 0.56 (0.44, 0.68) | 0.63 (0.51, 0.75) |
Fusion* radiomics model | |||
Decision tree | 0.79 (0.73, 0.85) | 0.43 (0.31, 0.55) | 0.52 (0.39, 0.64) |
Logistic regression | 0.71 (0.64, 0.79) | 0.72 (0.59, 0.84) | 0.68 (0.56, 0.81) |
Random forest | 1.00 (1.00, 1.00) | 0.54 (0.41, 0.66) | 0.59 (0.46, 0.72) |
Fusion radiomics model | |||
Decision tree | 0.84 (0.78, 0.89) | 0.60 (0.49, 0.72) | 0.51 (0.39, 0.64) |
Logistic regression | 0.82 (0.75, 0.88) | 0.77 (0.65, 0.87) | 0.78 (0.67, 0.89) |
Random forest | 1.00 (1.00, 1.00) | 0.56 (0.42, 0.68) | 0.67 (0.56, 0.78) |