Table 2. Performance of the different models in the training and validation sets.

Models’ or radiologists’ performance	AUC	ACC	Sensitivity	Cut-off	Specificity	PPV	NPV
LR	0.725 (0.716)	0.672 (0.660)	0.788 (0.652)	0.390	0.586 (0.667)	0.586 (0.600)	0.788 (0.714)
RF	0.800 (0.711)	0.672 (0.623)	0.558 (0.391)	0.378	0.757 (0.800)	0.630 (0.600)	0.697 (0.632)
SVM	0.733 (0.696)	0.664 (0.604)	0.481 (0.391)	0.409	0.800 (0.767)	0.641 (0.562)	0.675 (0.622)
DT	0.824 (0.695)	0.730 (0.679)	0.923 (0.826)	0.541	0.586 (0.567)	0.623 (0.594)	0.911 (0.810)
Bayes	0.715 (0.690)	0.680 (0.604)	0.519 (0.261)	0.267	0.800 (0.867)	0.659 (0.600)	0.691 (0.605)
KNN	0.816 (0.693)	0.697 (0.660)	0.654 (0.609)	0.400	0.729 (0.700)	0.642 (0.609)	0.739 (0.700)
Adaboost	0.705 (0.624)	0.730 (0.660)	0.538 (0.348)	0.474	0.871 (0.900)	0.757 (0.727)	0.718 (0.643)
Xgboost	0.823 (0.688)	0.836 (0.698)	0.731 (0.609)	0.423	0.914 (0.767)	0.864 (0.667)	0.821 (0.719)
GBDT	0.625 (0.510)	0.680 (0.566)	0.25 (0.087)	0.441	1.000 (0.933)	1.000 (0.500)	0.642 (0.571)
Radiomics + CNN	0.885 (0.812)	0.811 (0.774)	0.865 (0.826)	0.425	0.771 (0.733)	0.738 (0.704)	0.885 (0.846)
Radiologist1	–	0.757	0.739	–	0.768	0.675	0.819
Radiologist2	–	0.811	0.8	–	0.818	0.75	0.857
Radiologist3	–	0.789	0.753	–	0.8	0.725	0.822
3D CNN	0.874 (0.709)	0.862 (0.717)	0.786 (0.767)	0.495	0.962 (0.652)	0.965 (0.742)	0.773 (0.682)
nnU-Net	0.922 (0.835)	0.919 (0.830)	0.900 (0.800)	0.506	0.943 (0.870)	0.955 (0.889)	0.877 (0.769)

LR, logistic regression; RF, random forest; SVM, support vector machine; DT, decision tree; KNN, k-nearest neighbor; GBDT, gradient boosting decision tree; CNN, convolutional neural network; AUC, area under the curve; ACC, accuracy; PPV, positive predictive value; NPV, negative predictive value. Training set, in front of the brackets. Validation set, in brackets.