Skip to main content
. 2023 Aug 3;14:1138239. doi: 10.3389/fphys.2023.1138239

TABLE 2.

Performance of all fusion models.

Models Fusion phase Classifier Training cohort Validation cohort
AUC (95%CI) Accuracy Sensitivity Specificity AUC (95%CI) Accuracy Sensitivity Specificity
Fusion 1 delta2AP-T1 and delta2PVP-T1 LR 0.839 [0.767, 0.903] 0.750 0.576 0.939 0.751 [0.558, 0.917] 0.763 0.759 0.778
SVM 0.813 [0.734, 0.888] 0.758 0.606 0.924 0.789 [0.625, 0.929] 0.711 0.655 0.889
Fusion 2 delta2AP-T1 and delta2HBP-T1 LR 0.846 [0.770, 0.911] 0.811 0.742 0.894 0.812 [0.639, 0.943] 0.737 0.759 0.667
SVM 0.804 [0.721, 0.874] 0.758 0.697 0.833 0.774 [0.604, 0.922] 0.711 0.690 0.778
Fusion 3 delta2AP-T1 and delta2PVP-AP LR 0.789 [0.707, 0.861] 0.758 0.591 0.939 0.766 [0.622, 0.904] 0.684 0.621 0.889
SVM 0.760 [0.675, 0.840] 0.727 0.636 0.833 0.762 [0.588, 0.912] 0.632 0.621 0.667
Fusion 4 delta2PVP-T1 and delta2HBP-T1 LR 0.823 [0.749, 0.889] 0.765 0.667 0.879 0.701 [0.488, 0.897] 0.658 0.621 0.778
SVM 0.811 [0.739, 0.880] 0.735 0.727 0.758 0.609 [0.368, 0.828] 0.579 0.586 0.556
Fusion 5 delta2PVP-T1 and delta2PVP-AP LR 0.754 [0.673, 0.830] 0.720 0.576 0.879 0.766 [0.596, 0.916] 0.711 0.724 0.667
SVM 0.752 [0.672, 0.830] 0.712 0.773 0.667 0.818 [0.658, 0.958] 0.816 0.862 0.667
Fusion 6 delta2HBP-T1 and delta2PVP-AP LR 0.854 [0.789, 0.913] 0.788 0.818 0.773 0.778 [0.593, 0.936] 0.763 0.828 0.556
SVM 0.858 [0.790, 0.919] 0.811 0.803 0.833 0.808 [0.627, 0.952] 0.763 0.793 0.667
Fusion 7 delta2AP-T1 and delta2PVP-T1 and delta2HBP-T1 LR 0.882 [0.816, 0.931] 0.833 0.742 0.939 0.808 [0.627, 0.954] 0.763 0.724 0.889
SVM 0.906 [0.846, 0.960] 0.864 0.833 0.909 0.808 [0.581, 0.975] 0.816 0.793 0.889
Fusion 8 delta2AP-T1 and delta2PVP-T1 and delta2PVP-AP LR 0.804 [0.731, 0.875] 0.742 0.682 0.818 0.625 [0.441, 0.800] 0.605 0.655 0.444
SVM 0.778 [0.693, 0.866] 0.765 0.803 0.742 0.743 [0.562, 0.904] 0.711 0.759 0.556
Fusion 9 delta2AP-T1 and delta2HBP-T1 and delta2PVP-AP LR* 0.862 [0.795, 0.912] 0.795 0.758 0.848 0.851 [0.717, 0.959] 0.842 0.828 0.889
SVM 0.866 [0.796, 0.922] 0.818 0.864 0.788 0.659 [0.470, 0.844] 0.711 0.828 0.333
Fusion 10 delta2PVP-T1 and delta2HBP-T1 and delta2PVP-AP LR 0.805 [0.731, 0.873] 0.735 0.848 0.636 0.732 [0.535, 0.903] 0.711 0.828 0.333
SVM 0.830 [0.761, 0.895] 0.780 0.727 0.848 0.774 [0.559, 0.938] 0.763 0.793 0.667
Fusion 11 delta2AP-T1 and delta2PVP-T1 and delta2HBP-T1 and delta2PVP-AP LR 0.850 [0.785, 0.910] 0.780 0.712 0.864 0.720 [0.521, 0.890] 0.658 0.655 0.667
SVM 0.857 [0.791, 0.918] 0.818 0.788 0.864 0.816 [0.630, 0.954] 0.658 0.621 0.778
Fusion 12 delta3PVP-T1 and delta3HBP-T1) LR 0.866 [0.800, 0.922] 0.818 0.788 0.864 0.835 [0.697, 0.954] 0.789 0.828 0.667
SVM 0.860 [0.789, 0.927] 0.811 0.758 0.879 0.743 [0.567, 0.906] 0.763 0.793 0.667

*Models with the best comprehensive performance were used for the construction of the nomogram.

T1, non-contrast T1-weighted imaging; AP, arterial phase; PVP, portal venous phase; HBP, hepatobiliary phase; LR, logistic regression; SVM, support vector machine; AUC, area under the receiver operating characteristic curve.