Table 2 Diagnostic performance of various machine learning-based radiomics signatures
Model | Training cohort | Internal validation cohort | External validation cohort | Validation cohorts | ||||
|---|---|---|---|---|---|---|---|---|
AUC (95% CI) | ACC (%) | AUC (95% CI) | ACC (%) | AUC (95% CI) | ACC (%) | Average AUC | Average ACC (%) | |
KNN | 0.829 (0.753—0.904) | 77.55 | 0.778 (0.675—0.862) | 72.62 | 0.768 (0.653—0.860) | 76.06 | 0.773 | 74.34 |
SVM | 0.884 (0.831—0.925) | 80.61 | 0.860 (0.767—0.926) | 83.33 | 0.841 (0.734—0.917) | 83.10 | 0.851 | 83.22 |
LR | 0.870 (0.815—0.914) | 82.14 | 0.809 (0.708—0.886) | 79.76 | 0.744 (0.627—0.840) | 74.64 | 0.777 | 77.20 |
RF | 1.000 (1.000–1.000) | 100.0 | 0.642 (0.530—0.744) | 63.10 | 0.642 (0.519—0.752) | 63.38 | 0.642 | 63.24 |
LDA | 0.830 (0.770—0.880) | 75.00 | 0.782 (0.678—0.864) | 77.38 | 0.717 (0.598—0.818) | 73.24 | 0.749 | 75.31 |
NB | 0.769 (0.704—0.826) | 70.92 | 0.762 (0.657—0.848) | 75.00 | 0.691 (0.571—0.796) | 69.01 | 0.727 | 72.01 |
XGBoost | 0.926 (0.880—0.959) | 85.71 | 0.686 (0.575—0.783) | 69.05 | 0.643 (0.520—0.753) | 69.01 | 0.665 | 69.03 |